Numerical Investigation of Thermal Performance with Adaptive Terminal Devices for Cold Aisle Containment in Data Centers

Overview

The study starts from the problem that the adjustment scale of many airflow organisation methods is too large to support refined data-centre operation. The proposed device adapts cooling capacity to server power in the rack by adjusting deflector angles. A full-scale CFD model of a real Shanghai data hall was built: 21.5 × 20.3 × 4.9 m, seven rack rows, 10 CRAHs, 48U racks of 0.6 × 1.2 × 2.2 m, a 0.75 m plenum and 0.6 × 0.6 m perforated tiles with 35% opening ratio. It was validated with 21 K-type thermocouples (±0.1 K) at each of rack inlet and outlet in seven layers; measured CRAH outlet was 14 °C at 2.7 m/s.

The RNG k-epsilon model took 27.4% longer to compute than standard k-epsilon while differing in accuracy by no more than 6.7%, so standard k-epsilon was used, with perforated tiles as porous-jump boundaries. A grid study from 260,000 to 1,340,000 cells showed results stabilising at about 560,000 cells. Servers in the bottom 1st-2nd rack positions received very little air because the thick rack base frame created a vortex at the inlet.

The device adjusts per-server airflow between roughly 0.5 and 1.5 times the mean flow, and cut the maximum rack temperature by up to 4.2 K (for example rack F7 from 310.9 K to 306.7 K). To remove the same hot spot, lowering CRAH supply from 293 K to 291 K dropped COP from about 3.2 to 2.6; the device saved 20.1% of cooling energy versus that approach and 4.2% versus raising airflow 1.1 times. The spread of server exhaust temperatures narrowed.

Limits: only a CFD study, validated against a single rack inlet and outlet; the device itself was not tested in the field. Reliability, control latency and cost of moving deflectors were not assessed.

Key points

1. Modelled hall is 21.5 × 20.3 × 4.9 m with seven rack rows, 10 CRAHs, 48U racks of 0.6 × 1.2 × 2.2 m, a 0.75 m plenum and 0.6 × 0.6 m perforated tiles with 35% opening ratio.
2. RNG k-epsilon took 27.4% longer than standard k-epsilon while differing in accuracy by no more than 6.7%; standard k-epsilon was used and tiles were modelled as porous-jump boundaries.
3. A grid study at 260,000 to 1,340,000 cells showed results stabilising at about 560,000 cells.
4. Validation used 21 K-type thermocouples (±0.1 K) at each of rack inlet and outlet in seven layers; measured CRAH outlet was 14 °C at 2.7 m/s.
5. Servers in the bottom 1st-2nd positions received very little air because the thick rack base frame created a vortex at the inlet.
6. The device adjusts per-server airflow between roughly 0.5 and 1.5 times the mean flow and cut maximum rack temperature by up to 4.2 K (rack F7 310.9 K to 306.7 K).
7. To remove the same hot spot, lowering CRAH supply from 293 K to 291 K dropped COP from about 3.2 to 2.6; the device saved 20.1% of cooling energy versus that approach and 4.2% versus raising airflow 1.1 times.

Using it in design

• For hot-spot removal, before choosing lower supply temperature or higher airflow outright, compare the COP drop (3.2 to 2.6) and the energy difference (20.1% and 4.2%) with a rack-specific air distribution option.
• Check in rack layout and blanking design that details such as a thick rack base frame creating an inlet vortex starve the lowest servers.
• Add to operating scenarios that when server power distribution changes (high-power server in lower positions), local air distribution may need adjustment.

Who should read it

Cooling engineers assessing rack-specific air distribution and energy efficiency in cold-aisle-contained raised-floor halls.

Document contents

1. 1. Introduction
2. 2. Methodology (2.1 Adaptive Terminal Device for Air Supply, 2.2 Adjustment Strategy of Terminal Device)
3. 3. Numerical Simulation (3.1 Simulation Model: geometric model, governing equations and boundary conditions, grid independent study and validation experiments; 3.2 Numerical Investigation Setup)
4. 4. Results and Discussion (4.3 Energy Efficiency Analysis)
5. 5. Conclusions

Related standards

• ASHRAE TC 9.9 Thermal Guidelines for Data Processing Environments

Read the document

Publication details

Authors

Hongyin Chen; Dezhi Li; Songcen Wang; Tianheng Chen; Ming Zhong; Yi Ding; Ye Li; Xianxu Huo

Publisher

Buildings (MDPI AG)

Year

2023

Volume

13

Issue

2

Pages

268

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/buildings13020268

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation