Simulating Bladed Rotor Systems

I have had some opportunities to attend to few technical issues in rotordynamics in general and
probably none on bladed systems in particular. Even though we have robust fundamental capabilities
such as simulating cyclic symmetry conditions, centrifugal force loading for static procedures,
substructuring, periodic loading, modal evaluation and harmonic steady state analysis, generating
Campbell diagram plots and few additional capabilities, it is probably ok to say that legacy usage is more
in this domain. Customers use software such as DyRoBes, XL Rotor, ANSYS, Fluent and so on in the
turbines, compressors industries that are heavy on rotordynamics simulations.

I took some interest to investigate this area to understand some of these capabilities of Abaqus.
So let us consider a quarter cyclic symmetric model with 6-blades. The blade design is taken in this way
to keep it simple. Two cases are considered, one where there is no shroud connecting the blades and
another that has a shroud constraining the blades. Presently in Abaqus, the cyclic symmetry model can
be used for the plain vanilla frequency extraction analysis.

The models considered are shown in Figure 1, (a) being the case without the shroud (b) case
with shroud. These designs have as expected some similarity but a clear distinction in terms of the
modal response as evidenced in the eigenfrequency vs mode number plot shown later.

Figure 1: (a) 6-blade rotor Model (b) 6-blade rotor model with shroud

The analysis conducted on them to evaluate the behavior was:

1. Apply a centrifugal load equivalent to rpms between 5000 to 10000.

2. Mode extraction for cyclic symmetry case using *FREQUENCY type LANCZOS for both (a) and (b)
3. Study the effect of using NLGEOM in behavior of structure and on its eigenfrequencies.
4. Compare the eigenfrequency map for the unshrouded and shrouded cases
 5. Discuss in context of applying loads from nozzles (through vanes) a resultant modal
superposition approach to study the behavior for different numbers of vanes, blades, nodal
diameters and forcing harmonics.

Using NLGEOM in carrying forward prestresses

For the case of a centrifugal load equivalent to rpm of 10000 on the unshrouded blade model,
Figure 2 shows the effect of NLGEOM on the eigenfrequencies. This increase in the frequency content of
the structure is due to what is called as “stress stiffening”. This has a higher impact on the lower order
modes than on the higher order as evidenced by the percentage variation. These percentages can vary
quite significantly for other configurations and hence are not indicative of a general behavior.

9000 60
55
8000
50

Percentage Variation
7000 45
40
6000
Frequency

35
5000 30
4000 25
20
3000
15
2000 10
5
1000
0 Nlgeom Off
Nlgeom on
0 -5
PercVariation
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Mode Number
Figure 2: Effect of prestressing and not on the modal evaluation
Mode Shapes for Unshrouded bladed rotor system (cylic symmetric model quarter sector with 6
blades/sector)

Some of the typical mode shapes are shown in Figure 3, picking a typical profile in the ranges
which have close packed frequency values as in Figure 2. Note that this is the case without shroud and
after stress stiffening at 10000rpm.

1. (220Hz) 2. (543Hz) 3. (1245Hz)

4. (1452 Hz) 5. (1452 Hz) 6. (1457Hz)

7. (1459 Hz) 8. (1460 Hz) 9. (1656 Hz)

 10. (1854 Hz) 11. (3311 Hz)

12. (3815 Hz) 13. (4080 Hz) 14. (4111 Hz)

15. (6739 Hz) 16. (6754 Hz) 17. (7985 Hz)

Figure 3: Mode shapes (in brackets are the Frequencies in cycles/sec)

I am uploading an animation of the modes and mode shapes for the case of the shrouded blades
for brevity and for your detailed reference.
 Fundamentally, for a system with 24 blades, the maximum number of nodal diameters possible
is 12 (maximum nodal diameters possible is constant and determined as N/2 or (N-1)/2 depending on
even or odd numbers of blades respectively). Which means to say that the number of inflection points
(where blade movement or deflection direction changes) can be as many as 12 in the mode shapes. We
are observing mode shapes of the first 150 modes. Based on the understanding of nodal diameters and
above mode shapes, we can for example pick some of the mode images shown above and provide a
description in Table 1.

Mode Image Description of Mode

2 Zero nodal diameter mode

3 First nodal diameter mode
4 Second nodal diameter mode
7 Fourth nodal diameter mode
15, 16 Pure blade modes
17 Circular disk mode

Table 1: Mode profile description of few modes

Impact of centrifugal forces on Modal Behavior

Further, to evaluate the impact of the centrifugal load rate, the plot in Figure 4 indicates that for
the almost flat sets of mode numbers (close bunch of eigenfrequencies), for the higher load rate the
frequency value bumps up slightly, indicating impact of stress stiffening and the criticality in considering
the right load rate.
 9000

8000

7000

6000
Frequency (Hz)

5000

4000

3000

2000
No Shroud - Centrif 5000rpm
1000 No Shroud - Centrif 10000rpm

0
0 20 40 60 80 100 120 140
Mode Number
Figure 4: Impact of stress stiffening introduced by changing centrifugal load rate

Now, comparing the fundamental modal behavior of the bladed system with and without
shrouds, it can be seen that the sets of closely bunched modes in the unshrouded case is now slightly
diffused with smoother transitions in the mode frequencies. This is the key difference in the behavior
due to the additional restraints imposed by the shroud.
 9000
8000
7000
Frequency (Hz)

6000
5000
4000
3000
2000
No Shroud-Centrif-5000rpm
1000
With Shroud-Centrif-5000rpm
0
0 20 40 60 80 100 120 140
Mode Number

Figure 5: Variation in modal evaluation when using a shroud and not

It can be clearly seen that the key overlap frequency values are at the transitions at ~3300Hz and
twisting modes at ~6700Hz.

Campbell Diagram

In Rotordynamics, Campbell plot is a wildly popular plot used to identify resonance load
conditions. It is a plot between the running frequency and the eigenfrequencies. With Abaqus, we have
had a script that uses the modal participation factors to compute and plot the Campbell diagram based
on the eigenfrequencies from simulation. However to use the same to study bladed systems, we have to
encounter a few limitations.

Limitations

As of Abaqus 6.14, one limitation that exists when the plot script is to be applied to bladed
system is that it requires the modal outcomes from a COMPLEX FREQUENCY step, whereas a cyclic
symmetric interaction cannot be used with that procedure. As a result, one is required to run a full
model (not a symmetric part) to be able to obtain the Campbell diagram from that script. Moreover, to
excite particular nodal diameters in the frequency step, it is facilitated in Abaqus only when using cyclic
symmetry conditions. This constraint loop can act as deterrence when one is expecting to plot the SAFE
diagram which is a means of plotting the mode nodal diameter and the eigenfrequencies to identify if a
resonance condition obtained in the Campbell diagram will infact be something which induces true
resonance. This is a convenient way to interpreting results, where true resonance conditions as per SAFE
model are that, the forcing frequency should be same as rotating frequency and the forcing harmonics
be the same as the mode shape profile.

Future work

In near future, the effort is going to be to run a cyclic symmetric model with loads applied in
different harmonics to mimic the impact of nozzle jet through the diffuser vanes. This can be used to
study the bladed rotor system in more detail. In industrial setting on an actual workflow will involve
running a CFD analysis to obtain the forces (periodic) impacting the blades and then do a steady state
evaluation to obtain the stresses and peak deflections in the blades and root. This will serve as a
comprehensive workflow implementation for such bladed system rotordynamics.

Author Note

The observations in this writeup are made from my perspective and mine alone. To the best of
my knowledge I have tried to ensure there are no factual or observational discrepancies, but for any that
may appear, I request one to kindly bring it to my notice.