Professional Documents
Culture Documents
David L. Carlson
Principal Engineer
The Boeing Company
Seattle, Washington 98124-2207
Ted Yantis
Principal Engineer
The Boeing Company
Seattle, Washington 98124-2207
ABSTRACT
This paper presents the Boeing rotor-dynamics analysis procedures using MSC.Nastran. These analytical
procedures are typically used in analyses supporting airplane and propulsion system designs, but they are
applicable to any structure in which the rotating parts can be idealized as a collection of rigid disks and line
elements co-linear with an axis of rotation. Each rotating component is modeled separately in its own
superelement, which allows structures to be modeled with an arbitrary number of components rotating at
different speeds about different axes. The standard MSC.Nastran solutions for frequency response,
complex eigenvalue, and transient analyses have been enhanced by Boeing to include both synchronous
and non-synchronous gyroscopic effects. Gyroscopic matrices are used in the steady state solutions, and
are included in the mass and damping matrices depending on whether the solution is synchronous or nonsynchronous. Specific topics discussed in this paper are: gyroscopic matrices, equations of motion, and
damping representation. An example problem is presented which illustrates the application of the rotordynamics analysis procedures. This example is a representative finite element model of an engine mounted
on a wing. The accuracy of the nonlinear transient analysis technique is validated with test/analysis
correlation results for an actual propulsion system during the fan blade loss event.
Notation
Ip
Mx
My
&x
& y
oi
Ri
p
g
i
i
gj
S ij
frequency of mode i
structural damping coefficient for superelement j
percent of the total strain energy that superelement j contributes in mode i
[K]
(1 + ig)[ K ] + i[ K 4]
[K4]
[B ]
1.0 Introduction
In recent years, airplane certification requirements have been mandated requiring rotor-dynamics failure events to be
rigorously analyzed. In order to satisfy these new requirements, the scope and complexity of the finite element
models of the airplane and propulsion systems have increased significantly. Typical solutions performed at Boeing
are to identify engine critical speeds, and to ultimately design airplane/engine/nacelle structures which tolerate the
extreme conditions such as fan blade loss events.
The changes and additions to the standard MSC.Nastran solution DMAPs are referred to in this paper as The
Boeing Rotor Deliverables. These DMAPs are updated by Boeing whenever a new version of MSC.Nastran is
released. All procedures described in this paper have notations used by MSC.Software in their MSC.Nastran
Documentation. More details regarding the standard MSC.Nastran procedures and equations described herein can
be found in Chapter 9 of Reference (5). The Boeing Rotor Deliverables require that each rotor be placed in its own
superelement. The polar moment of inertias from the mass matrix are then used to build the gyro matrices.
2.0 Theory
The rotor-dynamic superelement analysis procedure can be summarized by the flowing steps:
Step 1 Superelement matrix generation and assembly
Step 2 Superelement dynamic reduction
Step 3 Develop and reduce the gyroscopic matrices
Step 4 Repeat steps 1-3 for each superelement
Step 5 Assemble and reduce the final analysis matrices
M x
0
=
M y
I p
I p &x
0 & y
(1)
The matrix gyro capability available with the The Boeing Rotor Deliverables, the rotational speed of rotor i is:
i = oi + Ri REF
(2)
The assumption in The Boeing Rotor Deliverables is that the spinning structure can be modeled as an assembly of
rigid rotating bodies aligned on a common spin axis and connected by flexible beam type elements . The individual
gyroscopic matrices can be assembled into superelement gyro matrices and to be reduced to the superelement
boundary. This process is similar to that used for the mass and damping matrices, and described in Section (9) of
Reference (5). The superelement gyroscopic forces can be described as:
(3)
Where,
oi
GYRO
The gyroscopic matrices, [BGYRO] and [GGYRO], for the residual structure can then be added to the damping and mass
matrices as appropriate for the type of analysis being done, as described below. For synchronous analyses such as
synchronous frequency response or complex modes, the reference speed, REF , is equal to the speed of excitation
(frequency response) or the frequency of the mode (complex eigenvalues), in both cases denoted by
In this
case, [BGYRO], the matrix of gyroscopic coefficients related to the constant speed term of Equation 2 ( oi ), is added
to the system damping matrix and the proportional part, [GGYRO], is added to the system mass matrix, as described in
Section 2.2. For non-synchronous analysis types, such as non-synchronous frequency response, complex modes, or
transient response, the reference speed REF , is constant. Hence, REF must be defined by the analyst, and the
term [BGYRO] +
REF [GGYRO] is added to the system damping matrix, as described in Section 2.2.
= REF ) or as synchronous
analyses, where the reference spin speed is assumed to be equal to the mode frequencies. Complex normal mode
frequencies and vectors are extracted by means of the following equation:
(4)
Where,
[ B ] = [ B] + [ B ] ,
[B ] = [B ] + [B ] + [G
[M ] = [M ] i[G ] ,
[ M] = [ M],
Synchronous
GYRO
GYRO
REF
GYRO
],
Asynchronous
Synchronous
GYRO
Asynchronous
The Boeing Rotor Deliverables has the capability to step through a table of reference speeds, performing a
complete asynchronous analysis for each speed. This capability enables the analyst to construct a Campbell diagram
which is a plot of complex frequencies versus rotation speed.
= REF ) or as
synchronous analyses, where the reference spin speed is assumed to be equal to the frequency of the forcing
function. The equation of motion for steady state frequency response analysis can be written as:
([K ] + i [B ] [M ]){U ( )} = {P ( )}
2
(5)
Damping and mass matrices in equations (4) and (5) are the same.
(6)
gyroscopic forces. These forces are functions of displacements or velocities from the previous time steps.
[B ]
which is the system damping matrix may also include gyroscopic terms for non-synchronous linear analyses. All
damping components such as structural damping, Rayleigh damping, or Direct Modal Damping as discussed in
Section 2.4 can be included in
[B ] = [B ] + Wg3 [K ] + W14 [K 4] + [B
GYRO
] + REF [GGYRO ] + [M ] + [K ]
(7)
Properties that explicitly vary with time, such as failure or breakage analysis, are most efficiently modeled by using
transfer functions or direct matrix input in order to directly add terms to the [K], [B], and [M] matrices. The Boeing
Rotor Deliverables allows this type of input to be changed for each residual structure subcase in SOL 129 which is
nonlinear transient analysis.
(8)
{ X } = [ ]{ q}
(9)
T
2
{q} = [ ]T {F (t )}
{&&} [ ] [ ][ ]{&}
I
q + C q +
(10)
(11)
Dii = i i
(12)
If a different structural damping coefficient is assumed for each superelement, one rational way to calculate the
equivalent overall modal damping coefficient for each mode would be to sum the superelement structural damping
coefficients, using the fraction of strain energy that each superelement contributes to the mode as weighting factors:
i =
j
S ij g j
(13)
To calculate the matrix [C] that corresponds to matrix [D], start with the identity:
[ D] = [ ] [ C][ ]
T
(14)
[ ] is mass normalized:
[ ] = [ ] [ M ]
Since
[ C] = [ M ][ ][ D][ ] [ M ]
T
(15)
This calculated matrix [C] can then be used directly in equation (8) or can be added to matrix
[ B ] in equation (6).
The disadvantages of Direct Modal Damping are that it creates a very full viscous damping matrix, and all the modes
of response must be calculated. These disadvantages can be circumvented somewhat by combining structural
damping with Direct Modal Damping. In order to combine Direct Modal Damping with structural damping, the normal
practice is to use modes only up to max , the frequency at which the structural damping gives the correct value.
The modal damping coefficients are then adjusted so that they have their full value at a modal frequency of 0.0 Hertz,
and decrease in a linear fashion to g=0.0 at frequency max . This is to account for the fact that the structural
damping effect increases linearly from g=0.0 at 0.0 Hertz to the correct value at frequency max . The effectiveness
of the structural damping will continue to increase for frequencies above
Direct Modal Damping can be combined to give the desired damping level as a function of the frequency content of
the structural responses. Assuming that equation (13) is used to calculate the overall damping for each mode, the
appropriate modal damping coefficients to account for the addition of structural damping with a coefficient of g max
can be calculated by:
S ij g j
i =
g max i
2
max
j
Where max is given by PARAM, W3 in MSC.Nastran.
(16)
Parameter N1SPD: Changing this parameter allows the rotor speeds used for any matrix gyroscopic effect
calculations to change in a piecewise fashion with time, to account for engine spool-down. Typically, the small highspeed rotors are modeled with matrix gyro using this piecewise linear time varying technique, which is much more
numerically efficient than the non-linear force technique. In order to capture the effect of continuously varying spin
speed accurately, the gyroscopic effects of the low speed rotor which is normally the larger rotor are usually modeled
with NOLINs or a combination of NOLINs and matrix gyro.
5. Test/Analysis Correlation
The fan blade loss test/analysis correlation studies are performed during the engine certification test to validate the
finite element model and analysis process. The analytical loads are computed based on a nonlinear transient analysis
of the integrated finite element model for the engine/test stand configuration. Figure 11 shows a schematic of the
strut hardware which integrates the engine/nacelle structure with the test-stand structure. Figures 12 show the strut
interface loads with the engine rear mount and the test stand structure. Results are non-dimensional mainly to protect
the proprietary information.
6. Conclusion
The Boeing rotor-dynamic analysis procedures with MSC.Nastran are presented in this paper. The standard
MSC.Nastran solutions for frequency response, complex eigenvalue, and transient analyses have been enhanced
with The Boeing Rotor Deliverables to include the gyroscopic effects. SOL 129 is enhanced to simulate engine
behavior during fan blade loss events. Analysis results for a fictitious engine model are presented. The accuracy of
the nonlinear transient analysis technique is validated with test/analysis correlation results for an actual propulsion
system during the fan blade loss event.
7. Acknowledgments
Authors would like to thank Dr. Ray Frick from Pratt & Whitney for providing example engine model.
8. References
(1) H Bedrossian, N. Veikos, Rotor-Disk System Gyroscopic Effect in MSC.Nastran Dynamic Solutions, Presented at
the 1982 MSC.Nastran Users Conference.
(2) D. Bella and M. Reymond, Eds., MSC.Nastran DMAP Module Dictionary, Version 68, The MacNeal Schwendler
Corporation, Los Angeles, CA, 1994.
(3) D. Herting, MSC.Nastran Advanced Dynamic Analysis Users Guide, The MacNeal Schwendler Corporation, Los
Angeles, CA, 1997.
(4) K. Kilroy, Ed., MSC.Nastran Quick Reference Guide, Version 70.5, The MacNeal Schwendler Corporation, Los
Angeles, CA, 1998.
(5) R. Lahey, M. Miller, and M. Reymond, Eds., MSC.Nastran Version 68 Reference Manual, The MacNeal
Schwendler Corporation, Los Angeles, CA, 1995.
(6) G. Sitton, MSC.Nastran Basic Dynamic Analysis Users Guide, The MacNeal Schwendler Corporation, Los
Angeles, CA, 1997.
Interface
Stiffness
Subcase 1 (K2PP=DMIG1)
Subcase 2 (K2PP=DMIG2)
Time
Figure 1
Effective
Damping
Desired Damping Level
= Structural + Modal
g
Structural
Damping
Adjusted
Modal Damping Level
Response Frequency
Figure 2
max
Figure 3
Figure 4
10
Figure 5
Figure 6
11
Figure 7
Figure 8
12
13
Figure10
Figure11
14
Figure 12 Test/analysis correlation of strut interface loads with engine rear mount and teststand
15
16