2 Type of Diffuser
Diffusers have an important role in deciding the stable operat-
ing range of a compressor. A vaneless diffuser (VLD) provides a
wide operating range, but the growing demand of a more efficient
compressor necessitates the use of a vaned diffuser. A vaned dif-
fuser (VD) will provide higher pressure rise and efficiency at
design flow coefficient, but the off-design performance is much
poorer than the vaneless diffuser. The choking flow rate of a
vaned diffuser is very small due to the formation of a geometric
throat (Fig. 1) at the diffuser vane. This can be avoided by reduc-
ing the number of vanes (in other words, reduce the solidity). As Fig. 2 Computational mesh for VD configuration
such, the low solidity vaned diffuser (LSVD) provides a better
operating range than the vaned diffuser and a better pressure re- coefficients, (/ ¼ 0.45 and 0.60). For VD configuration, the solu-
covery than a vaneless diffuser. The first reference to low solidity tion at / ¼ 0.60 was not obtained as this is the choking limit for
vaned diffuser was in a Japanese patent by Senoo [12]. Steady as this configuration. The radial gap is varied by keeping the vane
well as unsteady flow analysis in centrifugal compressor with low leading edge at R3 ¼ 1.05, 1.10, and 1.15.
solidity vane has been conducted by Yoshinaga et al. [13], Meshing of the computational domain is carried out using ICEM
Osborne and Sorokes [14], Sorokes and Welch [15], Hohlweg CFD. Tetrahedral elements are used for meshing the domain. Near
et al. [16], Amineni [17], Koumoutsos et al. [18], Kim et al. [19], the wall, meshing is carried out using prismatic layers to account
Siva Reddy et al. [20], Engeda [21], and Issac et al. [22]. These for the boundary layers (Fig. 2). Roache [23] identified grid-
works helped in assessing the effect of solidity and inlet distor- convergence studies are the most reliable technique for the quantifi-
tions on the performance of the compressor and also to assess the cation of numerical uncertainty. Grid independent studies for
relative performance. The amount of information available impeller and diffuser are done separately. Three different types of
regarding the interaction effects of LSVD is significantly lower grids are used. These grids are at three different levels Grid 1 for
than VD in the open literature. fine, Grid 2 for medium, and Grid 3 is for coarse mesh. Table 1
shows the grid independent study for impeller. Here, pressure ratios
P2/P01 and P02/P01 are compared for the three different grids. Grid
3 Computational Details 1 is taken as the reference and compared the pressure ratio obtained
For the present investigation, a low specific speed centrifugal with this grid to that of other two grids. The percentage differences
compressor with a 20 impeller blade is selected. The shape num- are shown in Table 1. The differences are considerably less for
ber of the machine is 0.084 and detailed specifications of this Grid 2 when compared to Grid 3. Hence Grid 2 is considered for
compressor test facility are given in Anish and Sitaram [11]. For the analysis.
the unsteady interaction study, two types of diffuser configura- In the case of diffuser, pressure recovery coefficient (cp) is
tions are selected. These are vaned diffuser (VD) and low solidity taken as the parameter for comparison between different grids.
vaned diffuser (LSVD). The number of diffuser vanes is 22 in VD Table 2 shows grid independent studies for VD. Here, Grid 1 is
and 11 in LSVD. The simulations are carried out at different flow taken as the reference. The percentage difference between Grid 2
coefficients. These are design flow coefficient (/ ¼ 0.34), one and Grid 1 is 0.064% only whereas between Grid 3 and Grid 1 it
below design flow coefficient (/ ¼ 0.23), and above design flow is 1.06%. Considering the fact that percentage difference is less
Elements cp % difference
Fig. 4 Pressure coefficient contours at the diffuser wall (comparison expt. with CFD)
solidity, the stall is delayed in LSVD. Similarly, at higher flow and the total pressure loss coefficient (wloss ) and are shown for
coefficients, variations are also observed in the mass averaged val- R3 ¼ 1.05 and R3 ¼ 1.15. The variations are quite large at closer
ues. At closer radial gaps, the fluctuations in the pressure ratio val- radial gaps, possibly due to the strong interaction with the impel-
ues become higher due to a stronger interaction between the ler blade. The pressure recovery varies due to the variations in the
impeller and diffuser. These unsteady variations in the stage pres- total pressure loss inside the diffuser. The maximum cp occurs
sure ratio can be due to the variations either in the impeller, or in
the diffuser, or both. A detailed analysis showed that the unsteady where the wloss has a minimum. This indicates that the unsteady
variations in the impeller are only marginal and major contribu- variation in the pressure recovery is not because of the variation
tion come from the diffuser. Hence, the unsteady performance in the dynamic head. In other words, the unsteadiness has little
variations of the diffuser are shown in the next section. effect in the energy transformation process inside the diffuser.
The peak value of cp occurs at different time steps for / ¼ 0.23
and 0.34 at close radial gaps. Whereas at R3 ¼ 1.15, the peak
4.2 Variation in the Diffuser Performance. The unsteady occurs at almost similar time steps, or, in other words, at similar
performance variation of the diffuser over one impeller blade pas- relative positions of the impeller and diffuser.
sage is plotted in the Figs. 6 and 7 for VD. The diffuser perform- Unsteady variation of cp and wloss for LSVD at different inter-
ance is evaluated in terms of the pressure recovery coefficient (cp) action levels is shown in Figs. 8 and 9. Unlike VD, in LSVD, two
Fig. 6 Variation of cp and wloss at different flow coefficients Fig. 7 Variation of cp and wloss at different flow coefficients
(VD and R3 5 1.05) (VD and R3 5 1.15)
peak values are observed for cp and wloss at two different time
steps. This is because two impeller blade passages are coupled
with a single diffuser vane passage in LSVD. The occurrence of
peak cp value corresponds to the blade passing frequency. At
/ ¼ 0.34, the unsteady variations in cp and wloss values are small
and at / ¼ 0.45, the variations are almost nil. Most of the static
pressure recovery occurs inside the vaned portion of the diffuser.
At / ¼ 0.60, the variations in cp and wloss increases again. Note
that, at this flow coefficient, the variations are not smooth, possi-
bly due to the large separations from the diffuser vane surface. As
in the case of VD, in LSVD, the peak cp value also occurs at
almost similar relative position of impeller and diffuser. It seems
that as the radial gap increases, the variation in the cp becomes
more deterministic. The physical mechanisms that control the
unsteady variation in cp is discussed in the coming section.
Fig. 12 Vortex core region within the diffuser passage (LSVD,
R3 5 1.10 and / 5 0.23)
4.3 Physical Explanation of Variation of Diffuser
Performance
vane leading edge, lower concave surface of the vane and trailing
Flow Field Analysis. Total pressure contours at two different edge of the vane. The strength of the vortex core region at the lead-
instances corresponding to the maximum and minimum value of ing edge is small when compared to the other two. The major
cp is shown for R3 ¼ 1.05 at / ¼ 0.23. For VD, the total pressure source of unsteady variations in the losses is found to be the trailing
contours are shown at s ¼ 0 and s ¼ 0.36 s at the midspan location edge vortex and the separation vortex from the lower surface of the
(Fig. 10). The loss regions are found to be located near the trailing diffuser vane. These two vortex regions grow in size and detach
edge of the diffuser vane. Similarly for LSVD, a majority of the from the diffuser vane and move to the exit. The vortices are shed
losses occur near the trailing edge of the diffuser and also from into the through flow at frequent intervals of time. Figure 13 shows
the lower concave side of the diffuser vane (Fig. 11). In a similar the limiting streamlines at the lower concave surface of the vane.
manner, at higher flow coefficients, the separated flow regions are The strength of the vortex core region at s ¼ 0.45 s is found to be
observed (not shown here) from the pressure surface side of dif- much higher than at s ¼ 0.18 s. Similarly, at higher flow coefficient
fuser vane. vortices are generated from the upper convex surface of the vane.
The losses are generated due to the formation and shedding of At / ¼ 0.60, the separation zone is observed throughout from the
vortices from the diffuser vane. The generation of vortex core upper surface of the vane and the vortices are generated from the
region from the diffuser vane surface and its propagation through diffuser vane leading edge itself (Fig. 14). The variation in the vor-
the diffuser passage is shown in Fig. 12 for LSVD at / ¼ 0.23. At tex core region as well as the shedding of vortices from the diffuser
this flow coefficient, the vortices are generated mainly from the vane surface may be triggered by the variation in the diffuser vane
Fig. 15 Diffuser vane loading at different time steps for LSVD; R3 5 1.10; / 5 0.23
Fig. 17 Circumferential variation of flow angle at the vane lead- Fig. 18 Span-wise variation of incidence angle (VD and
ing edge (VD and / 5 0.23) R3 5 1.05)
Fig. 20 Power spectral density (PSD) distributions of the static pressure fluctuations (LSVD,
R3 5 1.10 and / 5 0.34)
Fig. 21 Power spectral density (PSD) distributions of the static pressure fluctuations (LSVD,
R3 5 1.10 and / 5 0.23)
Fig. 22 Power spectral density (PSD) distributions of the static pressure fluctuations (VD,
R3 5 1.10 and / 5 0.34)