| |
By Mark Keating, Fluent Europe Ltd.
View the pdf of this article
Gas turbine installations commonly involve housing the gas generator
unit inside a confined acoustic enclosure, designed to
attenuate noise levels and provide ventilation for controlled
convective heat rejection and gas fuel leak dilution. Throughout the
European Union, the potential for fuel leakage means that such enclosures
must conform to the newly introduced ATEX Directives legislation
(from the French Atmospheres Explosibles) as described in HSE (Health
and Safety Executive) Guidance Note PM84.
Rendering of the gas turbine model created in GAMBIT, showing ventilation inlet
(blue) and exit (red) ports; access doors (orange) are superimposed as a visual guide
In the detection of gas leaks it is important to balance the ventilation
system so that leaks can be diluted and removed from the enclosure,
preventing build-up to explosive levels. Dilution should not be so
great, however, that the true leak magnitude is masked at the detection
point. The philosophy is that leaks that cannot be detected are diluted
and those that cannot be diluted are detected. PM84 declares limits
on the size of the 50%LEL (lower explosive limit) gas cloud before detection
as less than 0.1% of the net enclosure volume, where 100%LEL
represents the fuel/air mixture just rich enough to be ignitable. Additionally,
PM84 requires that the sensor alarm threshold be set at “ideally less
than 5%LEL and no more than 10%LEL.”
An iso-surface of the 50%LEL cloud at (above) 3%LEL detection (corresponding to
0.083% of net enclosure volume), and (below) 10%LEL detection (corresponding to
1.61% of net enclosure volume); the turbine casing and enclosure walls have been
removed for visualization
FLUENT was used in a recent case study to investigate the conformance
to the PM84 guidelines at one particular power plant site in the
UK. A representative model of the inside of the enclosure, including the
gas generator, intake, diffuser, pipework, and enclosure walls, was built
using GAMBIT. Roof extraction fans were used to establish a negative
gauge pressure inside the enclosure, and side wall inlet vents were used
for air supply. A tetrahedral mesh of approximately 3.1 million cells was
used, with cells concentrated around the fuel manifolds. The level of
heat rejected from the engine casing was calculated from a heat balance
across the engine, and applied to the casing surface as the driving
thermal boundary condition. In conjunction with site measurements
of internal pressure and temperature, FLUENT was used to establish confidence
in a baseline operating case prior to leak introduction.
For the safety analysis, a range of leak source flow rates, directions,
and locations were modeled in an attempt to create the largest leak
cloud possible. The leak flow rates used were based on the detector
alarm threshold levels in the 3-10%LEL range. The leaks were introduced
one at a time, as point sources in the worst possible locations – regions
with relatively high ventilation air residence time and low local air velocity.
A two-species mixture model (methane/air) was used to examine
diffusion of the gases and a user-defined function (UDF) was used to
calculate the sizes of the gas clouds.
The sizes of the steady state 50%LEL gas clouds for the worst leak
considered were predicted at the 3, 5 and 10%LEL detection levels. These
were determined to be 0.083% and 1.61% of the net enclosure volume
for the first and last cases, respectively. The sizes that these gas
clouds would reach before being detected depend on the exit detection
levels. In this instance, it was decided to install detection capable
of reaching the 3%LEL level rather than use a less sensitive limit coupled
with localized dilution or ventilation system modifications. Detection
at the 3%LEL level ensured compliance with the PM84 guidelines in
that the cloud size was <0.1% of the net enclosure volume.
|
|
|
