Professional Documents
Culture Documents
GT2017
June 26-30, 2017, Charlotte, NC, USA
GT2017-63126
ି
భ തതതത
భ
ܥൌ തതതത തതതത
భ ି ೞభ
4. NUMERICAL APPROACH
Table 1: Table of Uncertainties of Performance Variables
Numerical analysis of the scaled rig geometry was completed for
all the combinations of inlet swirl flow distributions as well as strut
angles and geometries. CD-adapco STAR-CCM+ 10.06.009 was used
for all mesh generation and numerical computations. CFD predictions
made with an earlier version of this code (v5.04) were previously made
by Bernier et al. [22] for diffuser-collector geometries operating with
uniform and axial inlet flow conditions and with incompressible flow.
The authors made comparisons between test data and predictions made
with the realizable two-layer K-Epsilon turbulence model and reported
CFD under-predicted pressure losses by 5%. In the present study, the
computational domain included the full test rig as shown in Figure 3.
All numerical models included the bellmouth, the swirl vanes, the
contraction, and the scaled diffuser-collector. The swirl distribution at
Figure 11: CFD Prediction capability of endwall static pressure recovery for select turbulence model
a) b) c) d)
Figure 12: CFD predictions of the Mach number through the exhaust diffuser-collector cross-section
a) K-Epsilon Two-Layer, b) SST, c) DES-EB, d) DES-SST
BACK WALL
c) CFD Prediction – k – Epsilon Two - Layer Model d) CFD Prediction – K – Epsilon EB Model
e) CFD Prediction - RSM Model – Linear Pressure Strain f) CFD Prediction – RSM Model – Quad. Pressure Strain
g) CFD Prediction – DES-EB Model – Ensemble Average h) CFD Prediction – DES-SST Model – Ensemble Average
Figure 13: PIV measurements and CFD predictions of collector exit normal velocity component normalized to inlet velocity
Figure 14: Exhaust diffuser-collector recovery change as a function of inlet swirl and strut stagger angle (top),
compared along the radial distribution of inlet swirl angle (bottom)
the influence of hub and shroud incidence on diffuser-collector recovery i. Strut stagger of +12º and inlet swirl angle of -14.6º. The resulting
was apparent when comparing the swirl angle sensitivity on the axial incidence angle along the hub and shroud were -34° and -19º
strut. A performance degradation was measured with the increase in respectively,
flow angle at the hub. Shroud incidence of the same magnitude did not ii. Strut stagger of -12º and swirl angle of -6.6º. The resulting
impact the diffuser-collector performance. For flows with radial incidence angle along the hub and shroud were +2° and +10°
distributions of inlet swirl, the influence of swirl on the performance of respectively, and,
this diffuser-collector exhaust system was dependent on the radial iii. Strut stagger of -12º and swirl angle of 9.4º. The resulting
location associated with maximum strut incidence angle. incidence angle along the hub and shroud were +15° and +34°
respectively.
The diffuser-collector recovery as a function of strut incidence
angle at the hub was plotted in Figure 15a and the shroud in Figure 15b. Distributions of relative recovery along the hub and shroud were
The data was provided in relative form consistent with the baseline shown in Figure 16 and Figure 17 respectively. The plots were
configuration with axial flow aligned with the axial strut. A definitive separated for data gathered at the top-dead-center (TDC) and bottom-
correlation of recovery with hub incidence was observed. Exhaust dead-center (BDC). The measurements were not gathered upstream of
performance did not correlate with shroud incidence. The accuracy of 38% normalized axial distance and the complementary CFD analysis
CFD performance predictions became increasingly unreliable as the was used in this range to provide insight in flow development within
flow angle deviated from the strut by 12º and higher. The accuracy of this region.
CFD predictions was also dependent on the radial location of incidence.
Accurate predictions were not achieved when large incidence angles at For the case with least incidence, an initial reduction in recovery
the hub were present. was observed within the strut commensurate to the reduction of cross-
sectional flow area due to strut blockage. Along the hub, gradual
The influence of strut incidence angle on flow development was diffusion within the struts was observed once the flow passed through
further evaluated with the use of static pressure measurements gathered the strut region of maximum thickness (maximum reduction of cross-
at the endwalls. These measurements were converted to a local recovery section flow area). Along the shroud, initial recovery was low due to
value and shown in relative form. The data was related to the total local flow acceleration at the diffuser inlet, where the axial flow-
exhaust recovery measured for the baseline case with no swirl and no conditioning flow path connected with the diffuser. For this low
incidence. Three scenarios were considered: incidence case, continuous diffusion was observed throughout the
length of the diffuser downstream of the strut mid chord. CFD
Flow visualization along both the hub and shroud walls was Similarly, the diffuser-collector geometry between cases a) and d)
completed to examine the impact of incidence on flow behavior near the were identical and the average swirl angle was increased from -6.6° to
endwalls. Photos of the streamlines along the hub and shroud were +9.4°. The increase in positive inlet swirl angle results in a stabilizing
shown in Figure 18 and Figure 20 respectively. The diffuser average effect on the flow along the hub despite an increase in strut incidence.
inlet swirl angle was set at -6.6°, and surface streak lines were visualized The flow interaction between the diffuser and collector were therefore
and compared for two cases of diffuser strut angle a) -12° (case ii above) sensitive to the diffuser inlet swirl angle. As seen in Figure 21 d), flow
and b) +12° (case i). As seen in Figure 18 a), separation along the BDC along the shroud separated at high incidence angles. The separated
of the hub was observed with the diffuser operating with little incidence. region shifted the general flow direction and resulted in an increase in
This separated region was at about 75% of axial length. The flow along the wetted area. The flow direction along the shroud followed the entire
the hub was completely changed by modifying the strut stagger angle, length of the diffuser. This was contrary to swirl angle sensitivity at the
and was characterized by recirculating flow along the entire diffuser hub, were significant strut incidence resulted in flow separation that
length at the BDC location. The transition from axial to radial flow that impeded effective utilization of the full diffuser length.
should occur within the collector commenced further upstream within
the diffuser. The reduction in exhaust diffuser-collector performance
resulted from the compounding effects of: 1) reduced diffuser length CONCLUSIONS
utilization that reduced the operating area ratio, 2) increased mixing
losses that resulted from flow separation and 3) increased blockage The impact of inlet swirl angle and associated radial distribution
caused by flow separation and increased through-flow velocities. The on diffuser-collector performance was examined along with the
streamlines along the shroud BDC as captured with flow visualization influence of strut stagger angle. Experimental measurements coupled
were shown in Figure 20. The flow along the shroud was mostly well with complimentary CFD analyses were completed on a quarter-scaled
behaved, and the flow direction was changed by the strut stagger angle. geometry representative of an industrial gas turbine exhaust. The
In Figure 20a) a region along the BDC existed were flow began to turn introduction of inlet swirl did not positively impact the performance of
well upstream of the diffuser-collector transition. The corresponding this diffuser-collector geometry. This finding was contrary to findings
strut incidence at the shroud was +10°. This flow phenomenon was not widely reported for annular diffusers without a downstream collector.
observed when the strut was re-set to +12º and the strut incidence at the The diffuser-collector was able to accommodate a significant amount of
shroud -14°. In summary, the flow along the shroud was changed with inlet swirl without a degradation in performance, so long as the inlet
a net increase in wetted area. The associated differences in flow flow direction did not significantly deviate from the strut stagger angle.
structure resulting from re-setting the diffuser strut were much larger A reduction in exhaust system performance was measured as the
along the hub. deviation between flow angle and strut stagger angle increased beyond
±12°. A symmetrical airfoil strut was used for this study and the impact
Comparative streamlines, as predicted with CFD analysis, were of strut geometry was not investigated.
shown in Figure 19 along the hub and Figure 21 along the shroud. These
velocity streamlines were plotted on a surface offset by 0.010 inches For flows with radial gradients of diffuser inlet swirl angle, the
from the endwalls. The CFD results in sub-labels a) and b) provided reduction in performance was dependent on the radial location
direct comparisons to streamlines shown using experimental techniques. corresponding to the highest strut incidence. Increases in incidence
Numerical predictions of streamlines were in good agreement with along the shroud had a weak effect on diffuser-collector performance.
experimental visualization in terms of capturing the overall flow Incidence effects on struts result in local flow separation that impact the
direction and associated strut impact. The flow along the hub was overall flow direction in the downstream section of the diffuser.
susceptible to separation at low strut incidence angles, and the full Changes in diffuser recovery were mainly attributed to differences in
length of the diffuser was not utilized along the BDC location. The wetted flow area and additional losses impacted by local flow
increase in strut incidence angle resulted in additional level of separation. A much stronger correlation between strut incidence at the
separation and a reduction in the utilized diffuser length. Predicted flow hub and diffuser-collector performance was identified. A reduction in
direction along the shroud was in good agreement with experimental diffuser-collector performance was measured as the swirl angle was
visualizations, and the impact of strut stagger angle on flow direction increased beyond 12º of strut incidence at the hub. Endwall static
was captured. CFD results showed an increase in flow velocities along pressure measurements, flow visualization, and numerical predictions
the shroud that resulted from flow separation induced blockage along were completed to identify differences in flow behavior at the hub and
the hub. shroud that impacted swirl angle sensitivities. Strut incidence at the hub
Figure 15: Impact of incidence at the (a) hub and (b) shroud on exhaust diffuser-collector recovery
Figure 16: Pressure recovery along the hub (a) TDC, (b) BDC
Figure 17: Pressure recovery along the shroud (a) TDC, (b) BDC
Figure 18: Surface flow visualization along the diffuser hub BDC location
a) Hub Incidence +2°, Shroud Incidence +10° b) Hub Incidence -22°, Shroud Incidence -14°
c) Hub Incidence -35°, Shroud Incidence -19° d) Hub Incidence +15°, Shroud Incidence +34°
Figure 19: Streamlines predicted with CFD along the diffuser-collector hub
Figure 20: Surface flow visualization along the diffuser shroud BDC location
a) Hub Incidence +2°, Shroud Incidence +10° b) Hub Incidence -22°, Shroud Incidence -14°
c) Hub Incidence -35°, Shroud Incidence -19° d) Hub Incidence +15°, Shroud Incidence +34°
Figure 21: Streamlines predicted with CFD along the diffuser-collector shroud