1. Experimental study of phase-change cooling and non-intrusive diagnostics for
data centre thermal management system
Vivek V. VYAS, A. Sridhar* and R. Narayanaswamy
Department of Mechanical Engineering, Curtin University, Perth, WA 6845, Australia
E-mail: R.Narayanaswamy@curtin.edu.au
ABSTRACT
As the demand for the data storage has been increasing day by day, the thermal designers
and server manufacturers are reviewing the current cooling techniques used in data centres
and are looking for immediate improvements. Direct immersion phase-change cooling is
one of the alternative cooling techniques which is more sustainable to the environment
and can replace air cooling currently used in data centres. The main focus of the study is
to isolate the effect of surface characteristics on the boiling curve and acoustic signature
of saturated pool boiling experiments. The present research comprises of the assembly of
test facility and the calibration of the equipment used. Furthermore, the experiments were
conducted using water as a working fluid at atmospheric pressure. One of the issues with
using direct immersion phase-change cooling is the sudden increase in temperature in
short interval of time at point of critical heat flux which can cause failure of the electronic
component due to overheating. The boiling acoustics have potential to be developed as a
non-intrusive diagnostic tool to identify the critical heat flux (CHF). Several tests were
conducted for pool temperature of 98°C to ensure repeatability, and the heating surface
temperatures and boiling acoustic emissions were measured for different heat fluxes. The
deterioration of the copper surface led to an increase in surface temperature for any heat
fluxes which deviated from following the expected boiling curves for saturated pool
boiling experiments. The boiling acoustics were obtained however it requires thorough
investigation to develop it further into the non-intrusive diagnostic tool for direct
immersion phase-change cooling techniques.
INTRODUCTION
The demand for data centres have been increasing in the past decade for data storage for the global
telecom and IT companies. This has led to the increase in energy consumption to operate the data centres
which has increased the carbon footprints and thus has resulted in global warming. It is estimated that
the computer room air cooling (CRAC) technique used currently takes about 40-45% of the total energy
consumed by a data centre to cool the electronic components (Marcinichen et al., 2011). It is very
important to ensure that the cooling technique used for cooling data centre is efficient and is more
sustainable to the environment. Direct immersion phase-change cooling offers opportunity to remove
high heat fluxes induced by convection on the surface of the substrate from high latent heat of
vaporisation of liquid coolant.
The aims of the current research involves in understanding the effect of surface characteristics on
boiling curves and acoustic signature of saturated pool boiling experiments. The other objectives
involved setting up the experimental test facility, perform
calibration tests for the measuring devices for accuracy and
conduct the saturated pool boiling experiments for water at
atmospheric pressure. The schematic of the experimental
facility can be seen in Fig. 1. The horizontal heating surface
used to obtain surface heat fluxes is copper and the
rectangular test chamber is made up of stainless steel. The
alterations were made in the test facility to measure the
boiling acoustics. The copper surface was prepared before
and after each tests, one of the remarks which can be made
from the present study is that the deterioration of copper
surface cannot be controlled while conducting the saturated
pool boiling experiments.
Figure 1: Schematic diagram of the
experimental facility
2. RESULTS AND DISCUSSION
The boiling curve of the preliminary subcooled test was validated using Rohsenow’s correlation
(Rohsenow and Bergles, 1964) which can be seen in Fig. 2 (a). There were four saturated pool boiling
tests conducted at atmospheric pressure. For the saturated tests, it was observed that the copper surface
was contaminated with a black layer due to the chemical reactions between the surface and dissolved
gas impurities of water followed by white calcium carbonate precipitate layer. The wall temperatures
of high contaminated copper surface were about 17°C higher than the low contaminated copper surface
in saturated tests which can be seen in Fig. 2 (b). The reason for this high variations was due to the
contaminated black layer formed on top of the copper surface which acted as an insulation and has a
lower thermal conductivity than copper. This resulted in increase of the wall temperatures of the heated
copper surface for all heat fluxes, therefore the low and high contaminated surface saturated tests were
incomplete to avoid melting of PEEK insulation on test section. The boiling curves and acoustic
signature of saturated tests were compared with different levels of contamination of the copper surface
to understand the effects of the surface characteristics.
The boiling acoustics were measured by an electret microphone and the frequency distribution of
each boiling regimes can be shown at different heat fluxes by using Discrete Fourier transform (DFT).
The DFT for two different heat fluxes which includes partial nucleate and fully developed nucleate
boiling for saturated tests were plotted as shown in Fig. 2 (c). It can be observed that the trend in
frequency distribution is similar at different heat fluxes, however the intensity wasn’t consistently
increased as per increase in the heat flux. The intensity of the sound was observed to reduce at high heat
fluxes which gives an indication that the CHF is approaching. The average frequency weighted intensity
for different contaminated surfaces for each heat flux was plotted in Fig. 2 (c) and it can be observed
that at low heat fluxes, the intensity of the sound produced by the high contaminated surface was highest
amongst other contaminated surfaces. From Fig. 2 (c), it can be seen that the average frequency
weighted intensity wasn’t increased significantly at higher heat fluxes.
CONCLUSIONS
The test facility assembly and calibration of the measuring devices were done successfully. The
methodology and the procedure followed for the experiments proved to be successful as the preliminary
subcooled test was validated with Rohsenow’s correlation. The results obtained from the saturated tests
were not accurate, however it can be used as a reference to understand the effects of surface
contamination and how it affects the boiling curves and acoustics associated. The further research is
required to understand effects of surface characteristics in order to develop a reliable non-intrusive
diagnostic tool and implement it in direct immersion phase-change cooling techniques in data centres.
REFERENCES
Marcinichen, Jackson. B., Jonathan A. Olivier and John R. Thome. 2011. Applied Thermal
Engineering (2012). doi:10.1016/j.applthermaleng.2011.12.008
Rohsenow, W.M. and A.E. Bergles. 1964. Trans. ASME, J. Heat Transfer. Vol. 86: 365-372
0
50
100
150
200
-20 -10 0 10 20 30
Heatflux(W/cm2)
Tw-Tsat (K)
Rohsenow's Correlation
Preliminary subcooled test
0
50
100
150
200
-20 -10 0 10 20 30 40 50
Tw-Tsat (K)
Rohsenow's Correlation
Low Contamination surface
Medium Contamination surface Test 1
Medium Contamination surface Test 2
High Contamination surface
0 50 100 150
-120
-80
-40
0
50 500 5000
Heat flux (W/cm2)
Soundlevel(dB)
Frequency (Hz)
Fully developed NB, q =130.15 W/cm²
Partial NB, q = 70.05 W/cm²
Low Contaminated surface
Medium Contaminated surface
High Contaminated surface
∆Tsub = 43.5°C
Working Fluid = Water
Pressure = 1.013 bar
∆Tsub = 2°C
Working Fluid = Water
Pressure = 1.013 bar
∆Tsub = 2°C
Working Fluid = Water
Pressure = 1.013 bar
Figure 2: Boiling curves of (a) clean copper surface, (b) for different copper surface contamination
level and (c) boiling acoustics of saturated pool boiling tests. This indicates CHF
(c)(b)(a)