A Conical Beam Finite Element for

L. M.Greenhill
Garrett Turbine Engine Company,
Rotor Dynamics Analysis
Phoenix, AZ The development of finite element formulations for use in rotor dynamics analysis
has been the subject of many recent publications. These works have included the
effects of rotatory inertia, gyroscopic moments, axial load, internal damping, and
W. B. Bickford shear deformation. However, for most closed-form solutions, the element geometry
has been limited to a cylindrical cross-section. This paper extends these previous
works by developing a closed-form expression including all of the above effects in a
H. D. Nelson linearly tapered conical cross-section element. Results are also given comparing the
Mechanical and Aerospace Engineering, formulation to previously published examples, to stepped cylinder representations
Arizona State University, of conical geometry, and to a general purpose finite element elasticity solution. The
Tempe, AZ 85287 elimination of numerical integration in the generation of the element matrices, and
the ability of the element to represent both conical and cylindrical geometries, make
this formulation particularly suited for use in rotor dynamic analysis computer
programs.

Introduction
Over the last several years, many investigators have ex- degrees of freedom per element. The Rouch and Kao element
tended the capability of rotor dynamics analysis using finite extended the earlier formulation by including gyroscopic
elements. Early formulations, such as Nelson and McVaugh effects and representing the area and inertia as second and
[1], developed a Rayleigh beam theory finite rotating shaft fourth order polynomials as a function of radius. The element
element which included the effects of translational and matrices were obtained by numerical integration, and as such,
rotatory inertia, gyroscopic moments, and axial load. This no closed form expressions were presented. It was also in-
development was subsequently generalized by Zorzi and dicated that the additional shear deformation coordinates
Nelson [2] to include the effects of internal viscous and could be condensed out of the element or system equations
hysteretic damping. Later, Nelson [3] added shear defor- without significant loss of accuracy. Later, To [7] developed
mation to the Rayleigh beam theory to develop a Timoshenko closed form polynomial mass and stiffness expressions for a
beam element, which was then extended by Ozguven and linearly tapered Timoshenko element, again using the twelve
Ozkan [4] to include the internal damping model of Zorzi and degree of freedom representation of Thomas, Wilson, and
Nelson. All of these formulations considered the axial cross- Wilson.
section of the element to be cylindrical, which allows area and This paper extends the linearly tapered Timoshenko beam
inertia to be considered constant with respect to length. The theory element by using the kinematic representation of
motion of these elements was represented by eight degrees of Thomas, Wilson, and Wilson to develop closed form
freedom: two translations and two rotations at the element polynomial expressions for element matrices suitable for use
ends. in finite element rotor dynamics computer programs. The
Modern rotor systems utilize geometry which is usually far element includes the effects of translational and rotatory
from being uniform as a function of length. These cross- inertia, gyroscopic moments, axial load, internal viscous and
sectional changes are mainly accommodated by modeling the hysteretic damping, and mass center eccentricity. System
rotor as a collection of stepped cylinders. For conical cross- equations of motion are also presented in both fixed and
sections, the errors introduced by the stepped cylinder ap- rotating reference frames. Numerical examples are given
proach may be quite large. comparing the use of the conical element to a previously
As a result, Rouch and Kao [5], developed a linearly published test case, to stepped cylinder representations, and to
tapered Timoshenko beam element for use in rotor dynamics. a general purpose finite element elasticity solution.
This element was based upon the work of Thomas, Wilson,
and Wilson [6], who determined that the most optimum Coordinates and Shape Functions
representation of the shear deformation was to add two
additional coordinates at each element end, resulting in twelve A typical axial cross-section of a linearly tapered finite
element is shown in Fig. 1. Each end of the element is
associated with an inner and outer radius, denoted by r and R,
Contributed by the Technical Committee on Vibration and Sound and with the subscripts / andy referring to the left (s = 0) and right
presented at the Design Engineering Technical Conference, Cincinnati, Ohio, (s = l) ends of the element, respectively. Defining a non-
September 10-13, 1985, of THE AMERICAN SocmTY OF MECHANICAL ENGINEERS.
Manuscript received at ASME Headquarters, June 11, 1985. Paper No.
dimensional position coordinate \, equal to the ratio s/l, the
85-DET-32. inner and outer radii may be expressed as

Journal of Vibration, Acoustics, Stress, and Reliability in Design OCTOBER 1985, Vol. 107 / 421
Copyright © 1985 by ASME
Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Z(w)

Fig. 1 Conical element axial cross section geometry

r = r , - ( l - t ) + ryf (1(a))
R = R,(l-Z)+Rj£ (1(b))
Representing the ratios of inner and outer radii on each end as Fig. 2 Kinematic relationships between element degrees of freedom
p and a, which are equal to rj/r,- and Rj/Ri respectively,
allows equations (1) to be rewritten
r = r,(l+(p-l)9 (2(a)) = 4[R,*(o- i ) - f . 4 ( p - i)]/(/j,4 - r , 4 )
R = *,(l+(ff-l)8 (2(b)) 6[*;4(a-l)2-//(p-l)2]/(*/4-
Using equations (2) in a cross-sectional area equation results 63 A[R,\a-\f--A' •r,*Q>- •l)3V(R,*-r,*)
in the following second order polynomial expression 54 • 4/
[Rfio-iy-rfto-iyvUif-r,*)
2 2 2
A = Tt(R -r )=Ai[l + a^ + a2^ ] (3) The primary coordinate reference system is illustrated in
where the coefficients are Fig. 2. The (XYZ) triad is a fixed reference with the X-axis
coinciding with the undeformed center line of the element.
A, = *Vt,2-r,2) Although not shown, the (xyz) triad is a rotating reference
a, = 2[Ri2(a-l)-ri2(p-l)]/(Ri2-ri2) with the *-axis coincident with X, and the y- and z-axes
rotating at a uniform rate to about the X-axis. The element is
a2 = [Ri2(a-\)2-ri2(p-^2V(Ri2-ri2)
considered to be initially straight and is modeled with twelve
Similarly for cross-sefctional inertia, the use of equations (2) degrees of freedom: two translations, two rotations, and two
results in a fourth order polynomial expression of shear deformations at each end-point of the element. The
diametral cross-section of the element is considered to be
/=7r( J R 4 -/- 4 )/4 = / / [ l + 5 1 ? + 6 2 ? 2 + S 3 ? 3 + 5 4 n (4)
circular.
where the coefficients are The translation of the element, neglecting axial motion, is
/, = T ( / ? , 4 - r , 4 ) / 4 given by the two displacements (v,w), and the shear defor-

Nomenclature
area polynomial coefficient k = transverse shear form factor (constant)
inertia polynomial coefficient P = element axial load (constant)
p,a element end inner and outer radii ratio t = time
V- mass per unit volume u,v,w = translational displacements in X, Y, and Z
e,f location of mass center directions
Q spin speed ip) = element end displacement vector, rotating frame
03 whirl speed coordinates
A whirl ratio, Q/co [q] = element end displacement vector, fixed frame
internal viscous, hysteretic damping coefficient coordinates
M internal damping matrix [Q) = element unbalance force vector
translation dependent rows of shape function bending stiffness matrix
matrix [KB] =
[Ks] = shear stiffness matrix
rotation dependent rows of shape function matrix axial load stiffness matrix
shear dependent rows of shape function matrix [KA] =
[x] circulation matrix
ratio of axial position to element length, s/l [Kc] =
[MT] = translational mass matrix
rotational displacements rotatory mass matrix
shear displacements [MR] =
[G] = gyroscopic effect matrix
s axial position along element transformation matrix
length of element [N] =
I [R] = fixed to whirl frame transformation matrix
r,R inner, outer radius of element end
A element area (function of axial position) Subscripts
I element inertia (function of axial position) ij = left and right element ends
E element elastic modulus (constant) 1,2, . . = particular term in a matrix or polynomial ex-
EP potential energy pression
EK kinetic energy
ED dissipation function Superscripts
G element shear modulus (constant) , . = position, time differentiation

422 / Vol. 107, OCTOBER 1985 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 generalized forces are required, and will be developed in
separate sections below.
rftf*

Potential Energy. Using Timoshenko beam theory, the

differential potential energy can initially be written as a
combination of elastic bending, shear, and axial load ex-
pressions, with the effects of internal damping on bending
strain energy to be included separately. In terms of the
element internal degrees of freedom, the potential energy is
2dEP=EI(6y'2 +dz,2)ds + kGA(Py2 + (3z2)ds
+ P(v'2 + w,2)ds (10)
Equation (10) may be integrated over the length of the element
r ORIGINAL POSITION to obtain an expression for potential energy in terms of three
discrete stiffness matrices as
- ^ X
L 2EP={q}T([KB] + [Ks] + [KA]){q} (11)
Fig. 3 Deformed element axial cross-section (X-Y plane) in which [KB] is the bending stiffness matrix, [Ks] the shear
stiffness matrix, and [KA] the axial stiffness matrix. The
individual matrices are obtained from equations (6, 7, 8) as
mation contribution is expressed as (fly,@z). The rotation of
the cross-section is defined by the relationships [KB] = \' EmWVW]ds (12(a))
6y = -w'-j3z (5(a)) Jo
6Z = v' + Py (5(b)) [K,] = \'okGA(Olx\Tlx\ds (12(b))
as shown in Fig. 2. Note that the sign of 6y is set negative in
order to obey kinematic constraints. The deformed shape of
an element in the X-Y plane is shown in Fig. 3, which [KA] = {'WW]*
JO
PWVmds (12(c))
illustrates the additional rotation of the cross-section due to Jo
shear. The components of each of the three stiffness matrices,
The translation of a point internal to the element relative to equations (12), are given in the Appendix. Each term in the
the end-point displacements is approximated by the relation bending and shear stiffness matrices, [KB] and [Ks], are
fourth and second order polynomials, respectively, due to the
v(s,t) variation of inertia and area with length, equations (3) and
= Ws)][g(t)) (6) (4). Axial load is assumed constant as a function of length,
w(s,t) thus each term in the axial stiffness matrix is a single value of
order zero.
with the components of [\p] given in the Appendix. The in-
dividual shape functions are obtained by assuming the
transverse displacement varies cubically as a function of Kinetic Energy. The kinetic energy expression is a
length. Using equations (5) and (6), and assuming the shear combination of translational and rotational components. In
deformation varies linearly with length, the rotation of the terms of the element internal degrees of freedom, the dif-
cross-section is given by the relation ferential kinetic energy may be written as
2dEK = ixA(i>2 + w2)ds + Id(6y2 + 6z2)ds+IpQ2ds-QIpdz6yds
6y(s,t) (13)
= [<«*)] (?(0) (7)
6z(s,t) Equation (13) may be integrated over the length of the element
to obtain the complete kinetic energy of the element. In terms
and the deformation of the element due to shear is expressed of discrete matrices, the energy expression may be represented
by by the following relation
2EK={q\T([MT} + [MR]){q\+n{q}T[GUq} (14)
Py(S,0
[X(s)]{q(t)} (8) in which [MT] is the translational mass matrix, [MR] the
0z(s,t) rotatory mass matrix, and [G] the gyroscopic matrix. The
individual matrices are defined for the purpose of forming the
with the components of [<j>] and [x] given in the Appendix. Lagrangian, and are obtained from equations (6, 7, 8) as
The order of the coordinates in the end-point displacement
vector was chosen as [MT] = j^MtfWnflds (15(A))
T 9
[q} = { vi,wheyi,dzi,Vj,Wj,e),j,dZj,l3),iMziAj^v 1 <>
in which the first eight terms are identical to those in [1-4], wR] = joV«)[*n*]<fr (15(b))
and the last four are the shear deformation contributions. The
order of coordinates in equation (9) was selected to facilitate
condensation of the shear deformation coordinates. [G] = j o 2fjJ(OmT[N][<P}ds (15(c))

with Id = nl(£), Ip=2nl(£), and [TV] a 2 x 2 skew symmetric

Element Equations transformation matrix as defined in the Appendix. The
The element equations can be derived through the use of components of each of the matrices in equation (15) are also
Lagrange's equations. As such, quadratic expressions for given in the Appendix. Each term is a second or fourth order
potential energy, kinetic energy, dissipative functions, and polynomial.

Journal of Vibration, Acoustics, Stress, and Reliability in Design OCTOBER 1985, Vol. 107/423

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Internal Damping - Bending Potential Energy and pressions (20). The unbalance force vector components ( Q c J
Dissipation Function. The incorporation of internal and (Q s } are given in the Appendix.
damping uses the same type of linear models as [2], It can be shown that the element matrices given in equations
Proceeding from the constitutive relationship between axial (12), (15), and (20) are equivalent to the same expressions
stress to axial strain, the differential bending energy and given in [3] by assuming a uniform cylindrical geometry and
dissipation functions for the Timoshenko element with in- reducing the matrices from a size of 12 x 12 to 8 x 8 through
ternal damping are defined as static condensation.
idEP = Em\qywvmw\+vv\mrw\\{q\ds(\(>(d))
System Equations
2dED = vvEm[q\TWVWm}ds (16(6))
with the components of [?j] given in the Appendix. These In the fixed reference frame (XYZ), the system equation of
expressions differ from those presented in [2] due to the motion may be assembled by the use of Lagrange's equations
relationship between bending and shear as given by equations from the expressions developed in equations (11), (14), (17),
(5). and (19). The resulting equation is of the form
Equation (16(a)) may be expanded and rewritten as {[MT] + [MR]){ q) + (r,v[KB] - Q[G]){ a]

2dEP =.£/({) , ' ' //+1 T

\q\ WYW\{q\ds + {" l'H + 1 [KB] + [KA]+ (r,vQ+ TM=)[Kc]]{q)
Jl + ito'
= {QF\ (22)
nV^{q}TW]T[N]T[<t>'}lq)ds All the matrices contained in equation (22) are symmetric
•fl + VH except for the gyroscopic matrix [G] and the circulation
and then may be integrated along with equation (16(6)) over matrix [Kc].
the length of the element to form matrix expressions for the For isotropic systems, it is convenient to use the rotating
bending potential energy and dissipation functions as reference frame (xyz). The transformation between fixed and
rotating frame coordinates is given by
2EP = | „!<?[,TlKB][g)
JT + VH lQ)=[R][p) (23)
VH T
where [R] is a transformation matrix whose components are
[ . 2+Qr„][g) [Kc]lg) (17(a)) given in the Appendix. Use of the transformation equation
+ VH (23) in the fixed frame equation of motion (22),
2ED = riv{q)T[KB]{q} (17(6)) premultiplying by [R]T, and defining a whirl ratio (A=fl/co),
gives the rotating or whirl frame equation of motion as
in which [Kc] is the internal damping circulation matrix,
obtained from equations (6, 7, 8) as
([MT] + [MR])[p] + { Vv[KB] + o([M r ] + (1 - X)[G] ] [p]
T T
[Kc] = ^oEm)[<t>'] [N] [4>']ds (18)
IK 1
VH
1(1+VHWB] + VH[KC]}
which is skew symmetric and each term is a fourth order + [KA]+Wly(.\-l)[Kc]
polynomial. The components of [Kc] are given in the Ap-
pendix. In equations (17), the matrix [KB] is the bending HlMT]+(l-2\)[MR])]{p} = lQc (24)
stiffness matrix given in equation (12(a)). Thus the in-
corporation of internal damping results in additional terms in in which the skew symmetric matrix [MT] is given by the
the potential energy expression as well as the creation of a expression
dissipative function. Note that if [0'] = [N] [\p"] as in [2], the
equations (16) could be transformed to the same expression [MT]=2\avAttmTlN\W>]ds (25)
for bending energy and dissipation functions as contained in
[2]. and the unbalance force vector [Qc] is given by equation
(20(a)).
Generalized Forces. The only generalized force included For a rotor composed of elements formulated in terms of
in this element formulation is due to distributed unbalance fixed reference coordinates by using equation (22) or in terms
force. For an element with mass center eccentricity (e(s), jfe)), of a whirl reference system through the use of equation (24),
the equivalent unbalance force is represented by the system will be composed of an assemblage of individual
lQF) = lQc}cosQt+{Qs}smQt (19) element matrices, each with size 12x12. This size may be
retained throughout the computation, or may be reduced
in which the force components are
using a static condensation algorithm as outlined in [1], at
either the element or system level.
[Qc) = fi Jo 2
[' M ( £ ) M r
| " I d s (20(a)) In this study, the elemental reduction approach was used,
Ito)J and each 12 x 12 element matrix was reduced to an equivalent
8 x 8 prior to assembly into the global system matrices. This
technique was evaluated in [5] and found to result in no
IQS) = Q2\ M « ) M r \ [ ds (20(6)) significant loss of accuracy. Computationally, this approach
Jo is identical to that used for subelement condensation in [1],
Ito)J and as such, the conical element may be treated with the same
using a linear distribution for mass unbalance over the reduction algorithm. The condensation of element matrices
element as prior to assembly in the global system matrices also facilitates
e(s) = e , ( l - £ ) + «,£ (21(a)) the addition of discrete support connections and lumped
masses.
to) = M l - 0 + fcf (21(6)) The calculation of whirl speeds or unbalance response uses
the equivalent unbalance force may be obtained from the same techniques as outlined in [1], and as such, will not be
equation (19) by using equations (21) in the integral ex- repeated. The addition of internal damping results in

424/Vol. 107, OCTOBER 1985 Transactions of the ASM E

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 Natural frequencies (Hz) of tapered chimney [6]
Five E ements Ten Elements Twenty Elements
Mode Thomas Current Thomas Current Thomas Current
1 0.5097 0.5092 0.5096 0.5092 0.5096 0.5092
2 2.6937 2.7024 2.6920 2.6928 2.6918 2.6904
3 6.5411 6.6725 6.5090 6.5454 6.5072 6.5128
4 11.355 11.800 11.073 11.262 11.062 11.105
5 17.748 17.583 16.070 16.627 16.019 16.162
6 24.109 33.988 21.338 22.540 21.162 21.501
7 38.902 41.310 26.882 28.941 26.381 27.039
8 40.756 46.413 32.790 35.605 31.589 32.700
9 45.333 49.652 38.963 42.040 36.712 38.380
10 49.554 51.736 40.607 43.209 40.380 41.192

STEPPED CYLINDERS

2.00 3.00 4.00 5.00 6.00

NUMBER OF SUBELEMENTS
Fig. 4 Typical stepped cylinder representation of conical geometry Fig. 5 Single element stepped cylinder versus cone stiffness com-
parison, term K ^
parametric terms in the rotating frame whirl speed
calculation, which can be accommodated through the use of
iteration procedures. In practice, these terms are usually
ignored and only undamped natural frequencies are
calculated in whirl frame analysis.

Numerical Examples
Three examples are provided to illustrate the accuracy and
use of the conical element. The first is a test case, originally
published by Thomas, et al. [6], in which conical elements
were used to calculate the natural frequencies of a tapered
chimney. Using five, ten, and twenty conical elements, the
results from [6] and those obtained from the present for- A - 60 OEG
mulation are compared in Table 1 for the first ten modes. ~I—
2.00 3.00 4.00 5.00 6.00 7.00
As the results indicate, the comparison is quite good, with
NUMBER OF SUBELEMENTS
the present formulation frequencies, in general, slightly
Fig. 6 Single element stepped cylinder versus cone stiffness com-
higher than those obtained in [6]. It was expected that the parison, term K 1 2
frequencies should differ, as the inertia and area represen-
tations used in [6] were linear rather than fourth and second
order polynomials as in equations (3) and (4). For reference,
the frequencies obtained with the present formulation are
practically identical to those presented in [5], being con-
sistently larger by a maximum of .35 percent.
The second example compares the use of the conical
element with stepped cylinders. As Fig. 4 illustrates, the
stepped cylinder representation is used with common rotor
dynamics programs to simulate conical structures. For the
comparison, the stiffness and natural frequencies were
evaluated for a cantilevered cone using from one to eight
steps, referred to as subelements as per [1]. Cone angles in-
vestigated were 0, 15, 30, 45, and 60 degrees, with the
thickness of the cones adjusted for angle, holding the normal 2.00 3.00 4.00 5.00 6.00 7.00 8.00
thickness to 1.0 cm and the mean radius at the fixed end to 2.0 NUMBER OF SUBELEMENTS
cm. Table 2 lists the properties of the cones used in this Fig. 7 Single element stepped cylinder versus cone stiffness com-
example. parison, term K22
For a single element cantilever cone, fixed at the small end,
the stiffness matrix can be reduced to a 2 x 2 matrix. The three drical formulation as in [3]. In Figs. 5, 6, and 7, the ratio of
terms in the resulting matrix may be compared for the stepped the stiffness term for the stepped cylinder (Ksc) to the conical
cylinder versus conical element formulations. Note that for element (Kc) is displayed for up to eight subelements. For
zero cone angle, the conical element degenerates to a cylin- terms Kn and Kn, translational and translational-rotational

Journal of Vibration, Acoustics, Stress, and Reliability in Design OCTOBER 1985, Vol. 107 / 425

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 K a
_ — _
X. q •

NESS RATIO, ^ v ^ \
.50
1 \ \

t 1 TERM
STI

* - K
^^~~-~~^*_———*
0 - K
O + - K
in .00 1.00 2.00 3.00 4.00 5.00 6.00
i i i i
—| 15.00 30.00 45.00 60.00 75.00
NUMBER OF SUBELEMENTS
CONE ANGLE, DEGREES
Fig. 10 One element stepped cylinder versus cone frequency com-
Fig. 8 Single beam element versus finite element elasticity stiffness parison, second mode
comparison

1.00 2.00 3.00

2.00 3.00 4.00 5.00
NUMBER OF SUBELEMENTS
NUMBER OF SUBELEMENTS
Fig. 11 Two element stepped cylinder versus cone frequency com-
Fig. 9 One element stepped cylinder versus cone frequency com- parison, first mode
parison, first mode

Table 2 Conical element test case

/ = 10.0 cm E = 20.0 xlO 1 0 N / m 2
3
V- = 8304 kg/m G = 7.6923x10 10 N / m 2
k = .900
Cone A iigle Rt n *J O
0 2.500 1.500 2.500 1.500
15 2.518 1.482 5.197 4.161
30 2.577 1.423 8.351 7.197
45 2.707 1.293 12.707 11.293
60 3.000 1.000 20.321 18.321

coupling, the comparison results in a family of curves, in-

creasing in value with increasing cone angle. The rotational
term K22 displays mixed results. In general, the comparison NUMBER OF SUBELEMENTS
indicates that there is no optimum number of subelements Fig. 12 Two element stepped cylinder versus cone frequency com-
that may be used to represent the conical geometry. parison, second mode
The stiffness terms predicted by the conical element were An additional comparison, that for natural frequencies of
also compared to those predicted by a general purpose finite the first two modes, is displayed in Figs. 9-12. In these
element computer code, ANSYS [8], The basic geometry as figures, the ratio of the natural frequency of the cantilevered
given in Table 2 was discretized into 80 axi-symmetric cone obtained by the use of stepped cylinders (cosc) to that
elements. Using cantilever boundary conditions, the resulting calculated by the conical element (coc) is presented for up to
displacements were used to calculate an equivalent 2x2 single eight subelements. Both one and two elements were used to
element stiffness matrix. The comparison of the terms in the calculate the frequencies, to illustrate the effect of refinement.
stiffness matrix is illustrated in Fig. 8 as the ratio of the beam As the figures illustrate, the ratio of natural frequencies is
element stiffness term (KBE) to that of the finite element poor unless several subelements are used. It is interesting to
(KFE). note that the second mode frequency is always overpredicted
As the figure indicates, the translational coefficient (Kn) using stepped cylinders. The comparison also indicates that
slightly underestimates the elasticity values. For the cross- no optimum number of stepped cylinders adequately predicts
coupling (K\i) and rotatory (K22) coefficients, the beam the natural frequencies of both modes for the cone angles
theory significantly overestimates the stiffness as given by the surveyed.
elasticity model, due to the inability of the beam theory to
accommodate any ovalizing effects for increasing cone angles.
It is somewhat surprising, however, that the deterioration in Summary and Conclusions
the agreement between the Kl2 and K22 terms is so immediate The equations of motion for a conical beam finite element
and rapid with any nonzero cone angle. have been developed from Timoshenko beam theory, and

426/Vol. 107, OCTOBER 1985 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 include the effects of translational and rotational inertia, Acknowledgment
gyroscopic moments, bending and shear deformation, axial
load, and internal damping. Both fixed and rotating frames of The authors gratefully acknowledge the programming and
reference system equations are presented. computational assistance of Mr. Wen-Jeng Chen in the
The resulting element formulation was compared to a preparation of this work.
previously published test case, to stepped cylinder
representations, and to a finite element elasticity solution.
Good agreement was obtained with the test case. For the
stepped cylinders, no optimum number of subelements was References
found to represent both stiffness and natural frequencies of 1 Nelson, H. D., and McVaugh, J. M., "The Dynamics of Rotor-Bearing
the cantilever cone example. The elasticity solution disclosed Systems Using Finite Elements," ASME Journal of Engineering for Industry,
the inability of the beam formulation to account for ovalizing Vol. 98, No. 2, May 1976, pp. 593-600.
effects for nonzero cone angles. 2 Zorzi, E. S., and Nelson, H. D., "Finite Element Simulation of Rotor-
Bearing Systems with Internal Damping," ASME Journal of Eng. for Power,
The conical geometry results in closed-form polynomial Vol. 99, No. 1, Jan. 1977, pp. 71-76.
expressions for the element matrices. These matrix 3 Nelson, H. D., "A Finite Rotating Shaft Element Using Timoshenko Beam
representations are easily incorporated into finite element Theory," ASME Journal of Mechanical Design, Vol. 102, No. 4, Oct. 1980, pp.
793-803.
rotor dynamics computer programs with minimal increases in 4 Ozguven, H. N., and Ozkan, Z. L., "Whirl Speeds and Unbalance
computation time and storage requirements. As the element Response of Multibearing Rotors Using Finite Element," ASME Paper
will degenerate to an equivalent cylinder, this formulation can 83-DET-89, Design and Production Technical Conference, Dearborn, MI, Sept.
replace rather than add to existing code. 11-14, 1983.
5 Rouch, K. E., and Kao, J. S., " A Tapered Beam Finite Element for Rotor
Use of the conical element greatly enhances modeling Dynamics Analysis," J. of Sound and Vibr., Vol. 66, No. 1, 1979, pp. 119-140.
flexibility. When confronted with conical geometry, the 6 Thomas, D. L., Wilson, J. M., and Wilson, R. R., "Timoshenko Beam
analyst is usually forced to represent the cross-section with a Finite Elements," J. of Sound and Vibr., Vol. 31, No. 3, 1973, pp. 315-330.
series of stepped cylinders or run independent conical section 7 To, C, W. S., "A Linearly Tapered Beam Finite Element Incorporating
Shear Deformation and Rotary Inertia for Vibration Analysis," J. of Sound
analyses. Use of the conical element eliminates this process, and Vibr., Vol. 78, No. 4, 1981, pp. 475-484.
resulting in a decrease in modeling time and improving the 8 DeSalvo, G. J., and Swanson, J. A., ANSYS Engineering Analysis System,
rotor representation. Revision 4.0.

APPENDIX
Translation Shape Function [I/-], equation (6)

h 0 0 ^2 ^3 0 0 1^4 -1/-2 0 -\j/A 0

M=
0 h ~*2 0 0 h ~*4 0 0 - ^ 2 0 -h
where
tf, = l-3$2+2$3
h = / ( £ - 2 £ 2 + ?3)
^3 = 3 £ 2 - 2 £ 3

+* = K-e+e)
Rotational Shape Function [</>], equation (7)
0 0i 0 0 -0i 03 0 "04 0 -04
[0] =
-0i 0 02 01 0 0 03 0
where
0, = m-e)
i-i = l - 4 £ + 3$2
03 = - 2 £ + 3£2
04 = 3 £ - 3 £ 2
Shear Deformation Shape Function [x], equation (8)

0 0 0 0 0 0 0 0 xi 0 x 2 0
M= 0 0 0 0 0 0 0 0 0 Xi 0 x2
where
Xi i-£
X2
General Notes on Presentation of Element Matrices
The next 9 element matrices are given in array format, that is, each nonzero term will be denoted by (ir, ic), where ir is the
row and ic the column in the matrix. This notation has been adopted for clarity, as each matrix term is a polynomial, except for

Journal of Vibration, Acoustics, Stress, and Reliability in Design OCTOBER 1985, Vol. 107/427

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 [KA], the axial stiffness matrix. All matrices are of size 12x 12, and any term not explicitly defined is zero. Since all of the
matrices are either symmetric or skew symmetric, only the lower triangular is given, with the type of symmetry given for each
matrix.
Element Bending Stiffness Matrix [KB], equation (12(a)) (sym)
K=EII/1U3
(1,1) = #(60480 + 302405,+241925 2 +211685 3 + 1900854)
(4,1) = A7(30240 + 100805, + 705652 + 60485-, + 547254)
(5,1) = (6,2) = - ( 2 , 2 ) = - ( 5 , 5 ) = - ( 6 , 6 ) = - ( 1 , 1 )
(8,1) = i:/(30240 + 201605i + 1713662 + 1512053 + 1353654)
(9.1) = -7^/(30240+ 151205, + 1209652 + 10584<53 + 95045 4 )
(11,1) = (10,2) = (12,2)= -(9,5)= -(11,5)= -(10,6)= -(12,6) = (9,1)
(3.2) = (5,4)= -(6,3)= -(4,1)
(7.2) = (8,5)=-(7,6)=-(8,1)
(3.3) = (4,4)=^/ 2 (20160 + 50405i+26885 2 +201653 + 172854)
(7,3) = (8,4) = tf/2(10080 +50405, + 43685 2 +40325 3 +37445 4 )
10,3) = tf/2(15120 +50405, +35285 2 + 30245 3 +27365 4 )
(12,3) = - ( 9 , 4 ) = -(11,4) = (10,3)
(7,7) = (8,8) = ^/ 2 (20160 +151205,+127685 2 + 110885 3 +97925 4 )
(10,7) = Kl2(\5\2Q + 100806, +85685 2 +75605 3 +67685 4 )
(12,7) = - (9,8) (11,8) = (10,7)
(9,9) = A72(15120 + 75605, +60485 2 +52925 3 +47526 4 )
(11,9) = (10,10) = (12,10) = (11,11) = (12,12) = (9,9)

Element Shear Stiffness Matrix [Ks], equation (12(b)) (sym)

K=klGA,/5\
(9,9) = (10,10) = i f ( 4 0 + 1 0 a , + 4 a 2 )
(11,9) = (12,10) = AT(20+10a,+6a 2 )
(11,11) = (12,12) = #(40 + 30a, + 24a 2 )

Element Axial Stiffness Matrix [KA], equation (12(c)) (sym)

K=P/30l

(1.1) = -(5,1) = (2,2) = -(6,2) = (5,5) = (6,6) = 36^

(4,1) = (8,1) = (6,3) = (9,5) = (11,5) = (7,6) = (10,6) == (12,6) = 3KI
(9,1) = (11,1) = (3,2) = (7,2) = (10,2) = (12,2) = (5,4) = -3KI
(3,3) = (10,3) = (4,4) = -(9,4) = -(11,4) = (7,7) = (12,7) == AKP-
(8,8) = (9,9) = (11,9) = (10,10) = (11,11) = (12,12) = AKP-
(9,8) = (11,8) = -(7,3) = -(12,3) = -(8,4) = -(10,7) = -(12,10) == KP

Element Translational Mass Matrix [MT], equation (15(a)) (sym)

M=ixA;l/9[
(1,1) = (2,2)=M(134784 + 31104a,+10944a 2 )
(4,1) = - ( 9 , 1 ) = - ( 3 , 2 ) = - ( 1 0 , 2 ) = M/(19008 +6048a,+2448a 2 )
(5,1) = (6,2) = M(46656 +23328a,+13248a 2 )
(11,1) = (7,2) = (12,2)=-(8,1) = M/(11232 +5184a,+2736a 2 )
(3,3) = M/ 2 (3456 + 1296a,+576a 2 )
(10,3) = ( 4 , 4 ) = - ( 9 , 4 ) = (9,9) = (10,10) = (3,3)
(6,3) = - (5,4) = (9,5) = (10,6) = -Ml(\ 1232 + 6048a, +3600a 2 )
(7,3) = -M/ 2 (2592+ 1296a, +720a 2 )
(12,3) = (8,4) (U,4) = (10,7) (9,8) = (11,9) = (12,10) = (7,3)
(5,5) = (6,6)=M(134784+103680a,+83520a 2 )
(8,5) = - ( 1 1 , 5 ) = - ( 7 , 6 ) = - ( 1 2 , 6 ) = - M / ( 1 9 0 0 8 +12960a,+9360a 2 )
(7,7) = M/ 2 (3456 + 2160a,+1440a 2 )
(12,7) = (8,8) = - (11,8) = (11,11) = (12,12) = (7,7)

428/Vol. 107, OCTOBER 1985 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Element Whirl Frame Translation Mass Matrix [MT], equation (25) (sk-sym)

M=pAiU9\
(2,1) = (6,5) =M(269568 +62208a, + 21888a 2 )
(3,1) = (10,1) = ( 4 , 2 ) = - ( 9 , 2 ) = - M / ( 3 8 0 1 6 + 1 2 0 9 6 a , + 4 8 9 6 a 2 )
(6,1) = -(5,2) = M(93312 + 46656a,+26496a 2 )
(7,1) = (12,1) = (8,2) = - ( 1 1 , 2 ) = A/7(22464 +10368a, + 5472a 2 )
(4,3)= - ( 9 , 3 ) = - ( 1 0 , 4 ) = (10,9) = M/2(6912 + 2592a,+1152a 2 )
(5,3) = (6,4) = - ( 1 0 , 5 ) = (9,6) = A//(22464 + 12096a, +7200a 2 )
(8.3) = (12,9)=-M/ 2 (5184 + 2592a,+1440a 2 )
(11,3) = M/ 2 (5184 + 2592a,+1440a 2 )
(7.4) = (12,4) = (9,7) = (10,8) = (11,10) = (11,3)
(7.5) = (12,5) = (8,6)=-(ll,6)=M/(38016 + 25920a,+18720a 2 )
(8,7) = -(11,7)= -(12,8) = (12,ll) = M/2(6912 + 4320a 1 +2880a 2 )
(10,9) = M/ 2 (6912 + 2592a,+1152a 2 )
Element Rotatory Mass Matrix [MR], equation (15(b)) (sym)

M=fiIi/9U
(1,1) = M(435456 + 2177285, + 12441652 + 77760S3 + 5184054)
(4,1) = M(36288 + 362885,+259205 2 + 181445 3 +129605 4 )
(5,1) = (6,2)= - ( 2 , 2 ) = - ( 5 , 5 ) = - ( 6 , 6 ) = -(1,1)
(8,1) = M/(36288 - 1036852 - 1296053 - 1296054)
(9.1) = -M/(217728 +1088645, +622085 2 +388806 3 +259206 4 )
(11,1) = (10.2) = (12,2): (9,5) (11,5) (10,6) (12,6) = (9,1)
(3.2) = -(6,3) = (5,4) = - ( 4 , 1 )
(7.2) = ( 8 , 5 ) = - ( 7 , 6 ) = - ( 8 , 1 )
(3.3) = (4,4) = M/2(48684 + 120966,+69126 2 +47526 3 +34566 4 )
(7,3) = (8,4)=-M/ 2 (12096 + 60485,+51845 2 +47525 3 +43205 4 )
(10,3) = M/ 2 (18144+181446,+129605 2 + 90725 3 +64805 4 )
(12,3) = (9,4)=-(11,4) = (10,3)
(7,7) = (8,8) = M/2(48384 + 362886, +311046 2 +280806 3 +259206 4 )
(10,7) = M7 2 (18144-51845 2 -64808 3 -64805 4 )
(12,7) = - ( 9 , 8 ) = - ( 1 1 , 8 ) = (10,7)
(9,9) = M/ 2 (108864+ 544326,+311046 2 + 1944063 + 1296064)
(11,9) = (10,10) = (12,10) = (11,11) = (12,12) = (9,9)

Element Gyroscopic Matrix [G], equation (15(c)) (sk-sym)

M=lili/9\l
(2,1) = M(870912 + 4354566, +2488325 2 +15552063 +1036805 4 )
(3,1) = -M/(72576 + 725765,+518405 2 +362888 3 +25920S 4 )
(6,1) = -(5,2)=-(6,5)=-(2,1)
(7.1) = A/7(- 72576 + 207366 2 +259206 3 +259206 4 )
(10,1) = -M/(435456 + 2177286,+1244166 2 +777606 3 +518406 4 )
(12,1) = - ( 9 , 2 ) = - ( 1 1 , 2 ) = - ( 1 0 , 5 ) = - ( 1 2 , 5 ) = (9,6) = (11,6) = (10,1)
(4.2) = (5,3) = (6,4) = (3,1)
(8.2) = - ( 7 , 5 ) = - ( 8 , 6 ) = (7,1)
(4.3) = M/2(96768 +241926, + 138246 2 +95046 3 +69125 4 )
(8,3) = - ( 7 , 4 ) = - M / 2 ( 2 4 1 9 2 +120965,+10368a 2 +95045 3 +86405 4 )
(9,3) = -M/ 2 (36288 + 362885,+259205 2 + 1814463 + 1296054)
(11,3) = (10,4) = (12,4) = (9,3)
(8,7) = M/ 2 (96768+ 725765, +622085 2 + 561605 3 +518405 4 )
(9,7) = M/ 2 (-36288+103685 2 + 1296053 + 1296054)

Journal of Vibration, Acoustics, Stress, and Reliability in Design OCTOBER 1985, Vol. 107/429

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 (11,7) = (10,8)-(12,8) = (9,7)
(10,9) = M/2(217728 +1088645,+622085 2 + 388805 3 +259205 4 )
(12,9) = -(11,10) = (12,11) = (10,9)
Element Circulation Matrix [Kc], equation (18) (sk-sym)
K=EIi/l\P
(2,1) = -#(60480 + 302405, + 241925 2 +21168S 3 +190085 4 )
(3,1) = #/(30240 + 100805, + 705652 + 60485-, + 547254)
(6,1) = -(5,2)=-(6,5)=-(2,1)
(7.1) = #(30240 + 201605,+171365 2 + 15120S3 + 1353654)
(10,1) = #/(30240+151205,+120965 2 + 105845 3 +95045 4 )
(12,1) = - ( 9 , 2 ) = - ( 1 1 , 2 ) = - ( 1 0 , 5 ) = - ( 1 2 , 5 ) = (9,6) = (11,6) = (10,1)
(4.2) = (5,3) = (6,4) = (3,1)
(8.2) = - ( 7 , 5 ) = - ( 8 , 6 ) = (7,1)
(4.3) = -#/ 2 (20160 + 50405,+26885 2 +20165 3 + 172854)
(8,3) = - ( 7 , 4 ) = -#/ 2 (10080 +50405, +43685 2 +40325 3 +37445 4 )
(9,3) = A72 (15120+ 50405, +35285 2 +30245 3 +27365 4 )
(11,3) = (10,4) = (12,4) = (9,3)
(8,7) = -#/ 2 (20160+151205,+127685 2 + 110885 3 +97925 4 )
(9,7) = A72(15120 + 100805, +85685 2 +75605 3 +67685 4 )
(11,7) = (10,8) = (12,8) = (9,7)
(10,9) = -#/ 2 (15120 +75605, +60485 2 + 52925 3 +47525 4 )
(12,9) = -(11,10) = (12,11) = (10,9)

Element Transformation Matrix [N], equations (15c, 18, and 25)

0 - 1 "
IN]--
1 0

Element Internal Damping Matrix [JJ], equation (16a)

M=
fl + VH
2
•1\H (1+1//)

Element Fixed to Whirl Frame Transformation Matrix [R], equation (23)

For the 12x 12 element matrices, [R] is a block diagonal matrix composed of six 2 x 2 matrices each equal to coswt[I) +
sino)t[N].

Element Unbalance Force Vectors ( Q c ) and ( Q s ) , equation (20)

M=Q2fiAi/ll
(Gci) = (Gsz) =A^e/(1764 + 420a 1 +156a 2 )+Afey(756 + 336a1+156cK2)
IQa) = -{Qsi}=A/r/(1764 + 420«,+156a 2 ) + Mf>(756 + 336a 1 +156a 2 )
lQa ) = -M/f,-(252 + 84a 1 + 36a 2 )-M/f,.(168 + 84a, + 48a 2 )
{Qc4 ) = M/e,(252 + 84a, + 36a 2 ) + M/e,-(168 + 84a, + 48a 2 )
IQa) = {Qs 6 }=Me / (756 + 420a 1 +264a 2 )+Afe i/ -(1764 + 1344a 1 +1080a 2 )
IQce) = - { e S 5 ) = M r / ( 7 5 6 + 420a,+264a 2 )+Mf,(1764+1344a,+ 1080a 2 )
[Qc) = M/f,(168 + 84a, + 48a 2 ) + M/f / (252+168a, +120a 2 )
\Qcs) = -M/e,(168 + 84a, + 48a 2 )-M/e,(252+168a, +120a 2 )
IQa) = (Q S 3 ) = (Gsio) = - ( Q c 4 ]
IGcio) = (Gs4l = - ( G s 9 ! = (ec3)
(Gcil) = {QS7) = IQS12) = - ( Q C 8 )
IQcn) = lQss} = -{Qsn) = lQa)

430/ Vol. 107, OCTOBER 1985 Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 01/27/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use