Journal of Sound and <ibration (2001) 240(4), 679}704

doi:10.1006/jsvi.2000.3263, available online at http://www.idealibrary.com on

ON THE EIGENFREQUENCIES OF A FLEXIBLE ARM

DRIVEN BY A FLEXIBLE SHAFT
O. KOPMAZR AND K. S. ANDERSON
Department of Aeronautical Engineering, Mechanical Engineering, and Mechanics,
Rensselaer Polytechnic Institute, ¹roy, N> 12180-3590, ;.S.A. E-mail: anderk5@rpi.edu

(Received 9 July 1999, and in ,nal form 20 July 2000)

The simulations of multibody dynamic systems with #exible components are generally
based on solving the equations of motion by using approximate methods. This approach is
taken because these systems' closed-form solutions are often not directly available. These
methods often assume a solution as a "nite series in terms of modal functions with
time-varying coe$cients. The eigenmodes of the system under study are preferable as the set
of the basis functions used in these series because such expansions provide greater accuracy
with fewer terms. As a consequence, accurate estimation of system eigenfrequencies and
eigenmodes is extremely useful (potentially necessary) in the e!ective modelling and
simulation of these systems. In this paper, a new general model consisting of rotor, shaft,
hub, beam, and payload, as might be encountered in certain industrial robots, is presented
and investigated. This model is similar in nature to those studied previously by a number of
researchers, but it is more general in form. The authors believe that this model contains
a more realistic (and higher "delity) representation of the rotor}shaft}hub assembly of this
system and its interaction with a #exible beam (arm) and associated payload. Through this
model the relative in#uence of seven key dimensionless mass, sti!ness and geometric
parameters (ratios) on system eigenfrequencies and modes may be qualitatively and
quantitatively investigated. These investigations may include many special cases such as
#exible shaft#rigid beam, rigid shaft##exible beam, cantilever}free beam, pinned}free
beam, "xed}free shaft, etc. Given the volume of numerical studies which may be performed
to this end, this paper concentrates on the e!ect of the two parameters representing the mass
and sti!ness ratios of the system manipulator on its driveline.
 2001 Academic Press

1. INTRODUCTION

Modelling and simulation of multibody dynamic systems has become indispensable in the
design and control of such systems. In many situations, these mechanical systems cannot be
adequately modelled and their behavior predicted if the #exibility, including #exible body
characteristics, of the system components are neglected. This is especially the case for
mechanical systems which have highly compliant or large links, and are subjected to high
driving torques in order to obtain high operation speeds. In such instances, the behavior
(and potentially performance) of the system can be heavily in#uenced by the link
#exibilities. In such cases, system modelling and analysis which includes all the structural
#exibility is helpful, if not absolutely necessary. Such modelling requirements may be

R Visiting Professor
Permanent Address: Department of Mechanical Engineering, College of Engineering and Archecture,
Uludag University, Bursa 16059, Turkey.

0022-460X/01/090679#26 $35.00/0  2001 Academic Press

680 O. KOPMAZ AND K. S. ANDERSON

unavoidable in order to develop reliable dynamic models which can be used to accurately
predict system response through the entire range of operation speeds, and scenarios. Such
a dynamic model is also useful in gaining insight into what design parameters most
in#uence speci"c aspects of the dynamic response of the system, and thus may be modi"ed
so to adjust the system behavior in a desired manner. In this way, one may determine which
parameters may be modi"ed so to most e!ectively reduce undesired vibrations. In
multibody dynamic analysis of continuous bodies, de#ections are usually approximated by
using truncated modal expansions. The use of a more accurate (representative) set of basis
functions allows for the use of fewer terms in the "nite series approximation, without loss of
accuracy. In this regard, a more accurate system model will help determine more realistic
eigenvalues and eigenmodes, which in turn may be used to produce a more accurate, less
complex discrete system model.
The #exibility of the drive train (driving shaft, transmission gears, etc.) of mechanical
system often plays an important role in the dynamic behavior of the system, and should be
included in the model. Typical examples of these applications are heavily loaded cam and
gear mechanisms, mechanical systems driven through long transmission shafts, some
spacecraft solar array drive assemblies and deployment mechanisms, and some robot
manipulators applications. The system studied in this work is in the form of
a motor}shaft}hub}one-ink robot arm}payload assembly, as shown in Figure 1. Through
the investigation of this system two principal objectives are pursued. Firstly, this work
studies the coupled e!ects of link and shaft #exibility on the eigenfrequencies of
a distributed parameter model (DPM) and compares the results with those of the models in
which the driving shaft or transmission line is represented by a torsional spring. This
comparison emphasizes the importance of modelling by demonstrating how signi,cant
variations in predicted behavior of the system can result from &&simple'' di!erences in the
model. The second aim of this work is to determine the exact eigenfrequencies leading to the
exact eigenmodes of the system, which may in turn be used to more e$ciently and accurate
solve the equation of motions by modal expansion.
The DPMs lead to a set of partial di!erential equations of motion. In general, an exact
and closed analytical solution cannot be obtained, thus some approximate solution
procedures must be applied. These procedures generally involve some form of model
discretization, which reduces the system of equations from a set of generally coupled
non-linear partial di!erential equations (PDEs), to a system of generally coupled
non-linear ordinary di!erential equations (ODEs). One of the most common discretizations
used in analytical dynamics is based on truncated ("nite) modal expansions. In this regard,

Figure 1. Full distributed system model.

 FLEXIBLE ARM ON SHAFT 681
the determination of the highly accurate, if not exact eigenmodes of the system (satisfying as
many boundary conditions as possible) is of great importance because it makes it possible
to obtain an accurate series solution with a reduced number of terms.
Since the late 1980s research associated with the two main scopes of this work have been
carried out. Barbieri and Ozguner [1] in 1988 investigated unconstrained and constrained
modal expansions for a rotating link. They concluded that pinned mode shapes are
inadequate for the analysis of #exible structures in slewing maneuvers. Furthermore, they
found that both the constrained and unconstrained mode expansions are not accurate in the
"rst #exible frequency as far as motor dynamics is not considered. In this work, the authors
did not account for joint or shaft #exibility.
Also in 1988, Yigi et al. [2] presented results considering a distributed radially rotating
beam with a tip mass, connected at the beam root to a rigid body. A set of coupled
non-linear equations were derived for the system using the extended Hamilton's principle.
The e!ect of coupling terms upon the vibration waveforms were investigated by using both
linearized and numerical solution of the di!erential equations. The results found that even
for small values of the ratio of the #exible beam to rigid-body inertias, the uncoupled
equations can lead to substantially incorrect results.
Bellezza et al. [3] derived the equations of motion of a rotating beam with a rotating
inertia at its base and a concentrated mass at its tip by using two di!erent non-inertial
rotating frames. They showed that the frequency equation does not di!er because the
physical system remains the same. These authors, however, do not consider the shaft or
transmission #exibility in general. A more detailed discussion of the resonance conditions of
deformable bodies and the use of multiple co-ordinate sets can be found in reference [4]. In
his paper, Shabana discusses and explains why the same solutions in multibody simulations
can be obtained even if two di!erent sets of mode shapes and two di!erent sets of
co-ordinate systems are used. He also emphasizes that the mode shapes and the co-ordinate
systems of deformable bodies undergoing large rigid-body displacements cannot be selected
arbitrarily.
In 1990, Low [5] analysed the eigenfrequencies of a Bernoulli}Euler beam attached to
a compliant, rotating hub. He developed a model in which the hub is connected to the base
by lumped torsional and linear springs. Furthermore, he de"ned some non-dimensional
parameter and studied the a!ects of various combinations including extreme cases (e.g.,
corresponding to di!erent boundary conditions of beam). The same author [6] later studied
the a!ects of hub inertia and payload on the vibration of a #exible*rotating beam. The
author states that the mode shapes are a function of the feedback control and the actual
frequencies depend on the motor inertias.
Garcia and Inman [7] presented the formulation of a model for a single-link #exible
beam, without end mass or payload, undergoing a slewing maneuver at an actively
controlled pinned end. The driveline sti!ness (motor, gears, and rotation position feedback
constant) was treated as an equivalent lumped linear spring, referred to as the &&servo
sti!ness''. They showed that for low- to-moderate ratios of the servosti!ness to beam
sti!ness, the treatment of the beam using clamped}free modes leads to an erroneous system
model.
Xi and Fenton [8] investigated the coupling e!ect of a #exible link and a #exible joint in
a one-link rotating structure. They de"ned two non-dimensional parameters (actually two
ratios: the ratio of a bending-type sti!ness of the link to the torsional sti!ness of the
rotor-beam joint; and the moment of inertia ratio of the link to the rotor) and compared
unconstrained (joint rotation a!ects link de#ection) and constrained (joint rotation does
not a!ect link de#ection) modal expansions. That same year, Xi et al. [9] studies the same
subject but for two special cases, i.e., a manipulator with a relatively #exible link and
682 O. KOPMAZ AND K. S. ANDERSON

a relatively rigid joint; and a second type with the properties contrary to the "rst. In the last
three works, the #exibility of the drive line, whether caused by the driving shaft or by the
shaft-to-link joint, is represented by a torsional spring. The detailed information on this
modelling in #exible joint robots is found in Potkonjak's paper [10].
In 1996, Morris and Taylor [11], using Hamilton's Principle, developed the equations of
motion for a single distributed #exible link, possessing a tip (payload) mass and rotational
inertia. The derivation presented in this work is very similar to that in reference [2], but
ignores the geometric sti!ness e!ects due to centripetal accelerations which were considered
by Yigit et al. [2]. This work demonstrated that the model presented by the authors
performed better in predicting experimentally determined results than the use of a
&&classical'' model.
The study presented in this paper di!ers from the aforementioned articles in that the shaft
is considered to be a continuous bar performing torsional vibrations. Additionally, the
coupled non-linear equations of motion are derived in very general fashion considering the
in#uence of rotor, shaft, hub, beam, and payload, as well as geometric sti!ness terms which
arise from both centripetal and Coriolis accelerations. Proceeding in this manner, a more
complete distributed parameter representation for the system is achieved which is felt by the
authors to be a more accurate (and thus exact) representation than the models presented in
the aforementioned works.

2. A COMPARATIVE MODEL STUDY

In this section, to emphasize the signi"cance of careful (more high "delity) modelling,
a comparison between a discrete parameter model (DcPM) and a DPM representations of
the same mechanical system is given. Consider the system shown in Figure 1 letting J be
"
the rotational inertia of a lumped rigid disc representing the hub, beam and payload which is
driven by the #exible shaft. o , G, I , A and ¸ indicates the density/length, the shear
Q Q Q Q
modulus, the polar second area moment, the cross-sectional area and the length of shaft
respectively. This is an unconstrained system with regard to rotational motion around the
z-axis and it may be modelled in two di!erent forms. The "rst of these forms is a DPM in
which the shaft has both distributed mass and distributed #exibility. In the second model
the #exibility of the shaft is replaced with a lumped torsional spring and an e!ective polar
inertia of the shaft is at one end of the spring. With k being de"ned as the ratio of the shaft
inertia J to that of the load (i.e., the lumped disc) J , the "rst exact frequency j , its
Q " 
approximation j from equation (A.12) and its percentage error are given in Table 1. From
this table, it is observed that the percentage error between j and j tends to zero for

k"&6...7, but increases quickly to '10% for k(5 or k'&9. Thus, even between very
simpli"ed models, the use of lumped inertia and sti!ness representations can give very
erroneous results.

3. A MORE EXACT DPM

The system model proposed in this work is the full system shown in Figure 1. In this
model the shaft is treated as a continuum connected to a rotor inertia, and is considered
separately from the distributed beam with lumped payload model, which are in turn
coupled through boundary conditions at the hub. Similar models has been used by other
authors [2}5, 8, 11, 13] in order to obtain the eigenfrequencies and eigenmodes of the
system, which in turn maybe used as basis functions in a "nite series for describing the
 FLEXIBLE ARM ON SHAFT 683
TABLE 1
j (DPM) and j (DcPM) values for di+erent k ratios

k j j Error"[(j !j) /j ] 100%
  
0)1 1)632 1)049 36
0)2 1)689 1)095 35
0)4 1)791 1)183 34
0)6 1)879 1)265 33
0)8 1)959 1)342 31
1 2)029 1)414 30
2 2)289 1)732 24
5 2)654 2)449 8
10 2)863 3)317 !16
15 2)948 4)000 !36
20 2)993 4)583 !53
25 3)021 5)099 !69

forced motion of the arm. Here, as a logical "rst step, we wish to con"ne our attention to
study the planar motion of the system, but otherwise the initial equations of motion derived
are very general in form with few simplifying assumptions.
The system most generally consists of "ve parts: rotor, shaft, hub, beam, and payload. In
most of the papers related to the subject, the hub inertia and hub radius are neglected for the
sake of simplicity. In other works, the payload is neglected (applicable for helicopter blades,
and many terrestrial manufacturing applications). In this derivation, the individual
contributions of the rotor, shaft, hub, arm and payload will all be considered.

4. EQUATIONS OF MOTION

For this system there are actually two equations of motion, which are coupled via
subsystem boundary conditions (BCs). The "rst of these represents the subsystem of the
rotor (motor armature, etc.) and shaft (driveline exclusive of motor and hub), while the
second represents the subsystem of the beam and payload (end e!ector plus workpiece).
These coupling subsystem BCs can be viewed as the equations describing the orientation of
the hub, which must be common to the equations of motion associated with each of these
subsystems. Figure 2 shows the reference frame, co-ordinate system, and basis vectors
associated with the hub}beam}payload subsystem, and with which the elastic deformations
of the beam are represented. The reference frame is "xed in the hub at the root of the beam
and is oriented such that the local x-axis of this rotating frame coincides with the
undeformed centerline of the beam.
The details related to the development of the equations of motion and to the derivation of
the eigenvalue (frequency) equation is given in Appendix B. In order to generalize the results
and aid in their comparison with those of other works, the follwing non-dimensional
parameter are introduced:
 o A¸ J EI /¸
R :"  " , R :" , (1, 2)
? J J @ GI /¸
0 0 Q Q
 o A¸ J  o A¸ J
R ":  " , R ":  " , (3, 4)
A o I¸ J B J J
Q Q Q Q & &
684 O. KOPMAZ AND K. S. ANDERSON

Figure 2. Local beam reference frame and co-ordinate system.

J M 1R
R ": ., R ": ., a"
: @, (5}7)
C 3J D M 3R
?

1R 1R r
b"
: @, c"
: @, rN "
: , (8}10)
3R 3R ¸
A B

where A represents the cross-sectional area of the beam, E represent the Young's
modulus of the beam material, I is the area moments of inertia of the beam, J , J , J , J
& . 0
are the rigid-body mass moment of inertias of the beam about its root, the hub, payload,
and rotor, respectively, ¸ is the length of the beam, M , M are the masses of the beam and
.
payload, o , o are the mass densities of the beam and shaft, respectively, and r is the hub
Q
radius.
The system, characteristic determinant, in terms of the above-mentioned parameters,
which leads to the eigenvalue (frequency) equation is given by equation (B.48) in
Appendix B, but its expansion is not given here due to its complexity. However, the
determinant given by equation (B.48) can be expanded by using any symbolic mathematical
code (e.g., Maple or Mathematica) as has been done for this study.
 FLEXIBLE ARM ON SHAFT 685
5. TWO SPECIAL CASES

In this section, the two special cases of the model presented in sections 3 and 4 are given.
These are the special cases of a #exible shaft}rigid beam system, and a rigid shaft}#exible
beam system, and represent limiting cases of R PR and R P0, respectively, for which
@ @
closed-form solutions exist.

5.1. FLEXIBLE SHAFT-RIGID BEAM

For this special (limiting) case, the system can be considered as shown in Figure 1, where
the beam is now treated is rigid. Here JI is equal to the summation of the hub moment of
inertia J , and the quantities J*, and J*, respectively, represent the beam and payload mass
& .
moments of inertia, each with respect to the rotation axis of the shaft. After some simple
calculus these quantities are found as

J*"(1#3rN #3rN ) J (11)

and

J*"J #(1#2rN #rN ) M ¸ . (12)

. . .
Hence,

JI "J #J*#J*"J #(1#3rN #3rN ) J #[J #(1#2rN #rN )M ¸]. (13)
& . & . .
The frequency equation of this now degenerate system is associated with a uniform shaft
which has two independent rotary inertias at each of its free ends, and performs torsional
vibrations as was discussed previously in section 2 (detailed in Appendix A) and is given as

g [(k k g!1) sin g!(k #k )g cos g]"0, (14)

   
where k and k are the inertia ratios de"ned as
 
J J /J 1/R R
k " 0" 0 " ?" A (15)
 J J /J 1/R R
Q Q A ?
and

JI J J J M ¸
k " " &#(1#3rN #3rN ) # .#(1#2rN #rN ) .
 J J J J J
Q Q Q Q Q
J J J J J M ¸ J
" & #(1#3rN #3rN ) # . #(1#2rN #rN ) .
J J J J J (M ¸) J
Q Q Q  Q
or

 
JI 1
k " " #(1#3rN #3rN )#3R #3 (1#2rN #rN ) R R (16)
 J R C D A
Q B
with g de"ned as

¸
g"u Q. (17)
cJ
686 O. KOPMAZ AND K. S. ANDERSON

TABLE 2
¹he eigenvalues obtained by equations (61) and (64) (R "1, R "10, R "R "10, rN "0);
? @ A B
conversion factor (b"18)2574

j (b j g
equation (B.48) equation (18) equation (14)

0)0278 0)0141 0)0141

0)4148 3)1417 3)1417
0)5866 6)2832 6)2832
0)7185 9)4248 9)4248
0)8296 12)5664 12)5664
0)9276 15)7080 15)7080

This case corresponds to the limiting case R PR. In the case of k "0 and/or k "0,
@  
equation (14) is the frequency equation of a free}free uniform bar with an inertia at one
end [9]. In our exact model, this case can be approached for R *10. For the parameter
@
values R "10, R "1, R "R "10, R "R "rN "0, the system eigenvalues j and g as
@ ? A B C D
obtained from equations (B.48) and (14), respectively, are listed in Table 2. Except at the true
limits represented by these special cases, the relationship

g"(bj (18)

exists between j and g, which arises because these two quantities are related to each other
via u.

5.2. RIGID SHAFT}FLEXIBLE BEAM

For this special case representing R "0, the system can be considered as a pinned}free
@
#exible beam with inertias "xed at the beam's ends. Here, JI includes J , J and J such that
0 Q &
JI "J #J #J . (19)
0 Q &
The frequency (or eigenvalue) equation is found to be

j [2j (rN #R !jR rN (R #rN )) sinh j sin j#2 (!1#jrN (rN #2R )
D C D D
#jR (R #2rN !jR rN )) sinh j cos j#(1!jrN (rN #2R )
C D D D
#jR (R #2rN !jR rN )) cosh j sin j #2j (R #R rN ) cosh j cos j
C D D C D
#k (j [(1#jR R )#(1!jR R ) cosh j cos j]
C D C D
#j [!(R #jR ) cosh j cos j#(R !jR ) cosh j sin j])]"0, (20)
D C D C
where

JI J #J #J J J J 1 1 1
k" " 0 Q &" 0# Q # &" # # . (21)
J J J J J R R R
? A B
 FLEXIBLE ARM ON SHAFT 687
For R "rN "R "R "0, equation (20) gives the frequency equation derived in reference
? C D
[8]. For the case where we additionally have k"0, equation (20) is the same as that of
a pinned}free beam. When kPR, equation (20) transforms into the frequency equation of
a cantilever beam provided that all the terms must be divided by k before k goes to in"nity
[12]. This case can be approximated with equation (B.48) by taking R su$ciently small
@
such that R )10\.
@
Note that the frequency equations related to these extreme (limiting) cases cannot be
obtained from equation (B.48) by the simple substitutions R PR or R "0, because there
@ @
are two di!erent continuous media in the model. However, the frequency equations of these
cases can be obtained from equations (B.32) to (B.41) by considering that A and B must be
zero for R "0, and similarly a, b, c, d"0 for R PR.
@ @

6. NUMERICAL RESULTS AND DISCUSSION

The frequency equation of the model presented in this work includes seven dimensionless
parameters: R , R , R , R , R , R , and rN . R , R , R , R , and R , are inertial ratios, R is
? @ A B C D ? A B C D @
a sti!ness ratio, and r is a geometric ratio. Through the resulting frequency equation (from
equations (B. 48)) it is possible through numerical studies to determine how each parameter
individually a!ects the eigenfrequencies. However, due to the incredible volume of possible
parameter value combinations that a thorough treatment of all seven parameters presents,
the authors have chosen to limit the results presented in this paper to the parameters R and
?
R . The in#uence of payload inertia ratios, R , R , as well as prescribed shaft input rotation
@ C D
rate u (0, t), are the topics of on going work.

6.1. EFFECT OF R
?
R is the ratio of the moment of inertia of the beam to that of the rotor, exclusive of the
?
shaft. A decrease in R means that either the beam inertia decreases or the rotor inertia
?
increases, i.e., a shift towards the case of a cantilever beam. It is reported in reference [1]
that R a!ects the frequencies and mode shapes of a system with a #exible link and
?
rigid joint. For this reason, four cases representing a wide range of R values (spanning
@
3 orders of magnitude) were simulated to investigate the e!ect of R on the system
?
frequencies. Figures 3(a}d), show the variation of the "rst four eigenvalues, respectively,
associated with all four R values considered. By comparison, Figures 4(a}d) show the
@
variations of the "rst "ve eigenfrequencies as a function of R , for each of the four chosen
?
values of R respectively.
@
Figures 3(a}d) indicate that in all cases the eigenfrequencies remain approximately
constant with R or increase only modestly in the "rst frequency. From inspection of
?
Figures 3(a}d) one concludes that the fundamental frequency is most sensitive to R .
?
Increasing R over the range of 0PR for a set R value represents a transition of the
? @
system from a pinned}free beam, with a spring opposing rotation of the beam at the pinned
end, to a pinned}free beam with no such spring. When R becomes larger the e!ect of R
@ ?
diminishes for all frequencies. Figure 3(b) shows the variation of the second frequency
versus R . For the cases of R '1 this frequency remains almost constant. A similar
? @
behavior is also observed for the third eigen for R '0)1, as indicated in Figure 3(c). Within
@
the range of R considered, the fourth and higher eigenfrequencies remain unchanged
@
irrespective of R ; Figure 3(d). Figures 4(a}d) demonstrate how the "rst "ve frequencies vary
?
on R for constant R . In all these "gures it is observed that only the "rst frequency varies
? @
688 O. KOPMAZ AND K. S. ANDERSON

prominently with variations in R , with the second eigenvalue varying almost

?
indistinguishably, and all higher eigenvalues remaining e!ectively constant.
Another interesting point to be mentioned is the decrease in all frequencies as R
@
increases. An increase in R corresponds to either an increase in the beam sti!ness,
@
a decrease in the shaft sti!ness, or a combination thereof. In any regard, an increasing R
@
can be interpreted that the shaft is becoming more #exible relative to the beam. Thus, it is
clear that the shaft plays a determining role in the formation of low frequencies. A similar
situation is observed in the graphics given in reference [7].

Figure 3. Variation of each of the "rst four eigenvalues with R for "xed R values: *, R "0)01; *, R "0)1; ;,
? @ @ @
R "1; #, R "10. (R "R "1000, rN "0)001.) (a) Eigenvalue C1, (b) C2, (c) C3, (d) C4.
@ @ A B
 FLEXIBLE ARM ON SHAFT 689

Figure 3. Continued.

It should be remembered that an increase in R can be achieved in a variety of ways (e.g.,

@
through the variations of E, G, I , I , ¸ and ¸ ). For instance, provided that all the other
Q Q
parameters within the de"nition of R are held constant, making the length of the beam
@
smaller yields a larger EI/¸ ratio, and consequently a large value for R . But this length
@
variation also has an e!ect on the other inertial parameters, because of J . For example,
@
a decrease of 5% in the beam length leads to a decrease of 14% in the beam inertia, which
causes variations of the same percentage in the R , R , R , R , and R parameters. Therefore,
? A B C D
one has to be careful when making such considerations. In a similar way, some care should
be exercised when considering the parameters rN and R because these two ratios include
B
quantities associated with the hub.
690 O. KOPMAZ AND K. S. ANDERSON

6.2. EFFECT OF R
@
R is the ratio of a form of beam sti!ness to that of the torsional sti!ness of the shaft. Here,
@
the shaft sti!ness is de"ned as the equivalent sti!ness of a uniform circular cross-section
shaft, i.e., GI /¸ . Figures 5(a}d) demonstrate how R a!ects the eigenfrequencies for four
N Q @
di!erent R values. Here it is observed that all eigenfrequencies decrease as R increases.
? @
One may view the increase in R as a decrease in the shaft sti!ness relative to the beam
@
sti!ness. If all other quantities are held constant, this would result in a drop in the shaft

Figure 4. Variation of the "rst "ve eigenvalues with R for di!erent R constants: *, eigenvalue C1; *,
? @
eigenvalue C2; ;, eigenvalue C3; #, eigenvalue C4; 䊉, eigenvalue C5. (R "R "1000, rN "0)001.)
A B
(a) R "0)01, (b) 0)10, (c) 1)0, (d) 10)0.
@
 FLEXIBLE ARM ON SHAFT 691

Figure 4. Continued.

torsional modes with an increase in R . This decrease in eigenfrequencies with increases in

@
R is shown in Figures 5(a}d).
@
The e!ect of R is most signi"cant in the fundamental mode and decrease with both
?
increasing mode number and R , becoming negligible for modes four and higher. When R
@ ?
increases so do all frequencies, again with the e!ect being negligible for all modes higher
than three.
Figures 6(a}d) show how all these frequencies behave for given changes in R and
@
constant R . A general decrease for all frequencies is observed, as has been pointed out
?
earlier, but a more interesting observation may be made from these "gures. Speci"cally,
drastic decreases in the fourth and "fth frequencies can be seen with a existence of
692 O. KOPMAZ AND K. S. ANDERSON

a well-de"ned elbow. This result represents a curve veering phenomenon, as discussed in

detail by Perkins and Mote [13]. In that work the authors showed that a curve veer, as
opposed to a cross, could result as an artifact of a discretization procedure used in the
problem, or solution approximation method such as a perturbation approach. In this work,
this curve veering is not caused by any numerical problems because no discretization
procedure or perturbation technique has been applied to obtain the frequencies, as it is
emphasized in that paper [13]. Therefore, this curve veering represents an actual event or
crossing of modes. Speci"cally, after the veering region the eigenmode shapes associated

Figure 5. Variation of each of the "rst four eigenvalues with R for "xed R values: *, R "0)01; *, R "0)1; ;,
@ ? ? ?
R "1; #, R "10. (R "R "1000, rN "0)001.) (a) Eigenvalue C1, (b) C2, (c) C3, (d) C4.
? ? A B
 FLEXIBLE ARM ON SHAFT 693

Figure 5. Continued.

with these frequencies interchange. This phenomenon cannot be observed in reference [8]
because only the "rst three frequencies were studied there.
When one compares Figures 3(a}d) and 5(a}d) of this paper with the results given in
reference [8], signi"cant di!erences are obvious. For example, a jump phenomenon of the
frequency curves does not appear in the graphics obtained in reference [8]. Moreover, the
frequency equation given in reference [8] does not yield the frequencies listed in Tables
2 and 3 of that article, indicating some form of error on the part of that article. By
comparison, the model presented within this work is felt to be more exact than the models
which are associated with related systems and have previously appeared in the relevant
literature. The frequency equation as derived in this work from equation (B.48) is very
694 O. KOPMAZ AND K. S. ANDERSON

general in scope, satis"es limit cases which have been checked against independently
obtained results of other authors, and gives reasonable and reliable results overall.

6.3. COMMENTS REGARDING R , R , R , R , rN , AND


A B C D
In the results presented in the preceeding sections of this paper, the e!ects of R
?
(and R ) for di!erent values of R (and R ) with the values of R , R , R , R , rN held
@ @ ? A B C D

Figure 6. Variation of the "rst "ve eigenvalues with R for di!erent R constants: *, eigenvalue C1;
@ ?
*, eigenvalue C2; ;, eigenvalues C3; #, eigenvalue C4; 䊉, eigenvalue C5. (R "R "1000, rN "0)001.)
A B
(a) R "0)01, (b) 0)10, (c) 1)0, (d) 10)0.
?
 FLEXIBLE ARM ON SHAFT 695

Figure 6. Continued.

constant, and
Q neglected, were investigated. To aid in the comparison of these results with
those obtained by other investigators, the parameter values of R "R "10, R "R "0,
A B C D
rN "10\, and
Q :0 were used. These parameter values represent a system with negligible
shaft, hub and payload inertias, as well as negligible sti!ening (softening) e!ects arising from
centripetal and Coriolis accelerations. It should be noted that the e!ect of these parameters
on the system eigenvalues and associated mode shapes are not in general negligible. Indeed,
preliminary work indicates that these parameters (individually, and in combination) can
signi"cantly alter the dynamic behavior of the system. This is clearly demonstrated by the
results shown in Tables 3}5, which indicate how the "rst four eigenvalues can be in#uenced
by changes in the dimensionless parameters R , R and rN respectively. Thus, the particular
A B
simplifying assumptions used in modelling the dynamic characteristics of the system need to
696 O. KOPMAZ AND K. S. ANDERSON

TABLE 3
¹he ,rst four eigenvalues for various R values (R "0)3, R "0)1, R "3, rN "0)
A ? @ B
R j j j j
A    
10\ 0)2272 0)2624 0)2933 0)3213
10\ 0)7358 1)0405 1)2743 1)4711
10\ 1)2784 1)7737 1)9196 2)2440
10\ 1)7628 2)1638 2)9854 3)7120
1 1)8215 2)9002 4)4882 4)8993

TABLE 4
¹he ,rst four eigenvalues for various R values (R "0)3, R "0)1, R "1, rN "0)
B ? @ A
R j j j j
B    
10\ 1)6922 1)8757 4)2090 4)6941
10\ 1)7066 1)8811 4)2107 4)6946
10\ 1)7831 1)9675 4)2280 4)6989
1 1)8191 2)5620 4)3636 4)7523
10 1)8223 3)1053 4)5500 5)1228
100 1)8226 3)2029 4)5706 5)2743
1000 1)8226 3)2132 4)5725 5)2907

TABLE 5
¹he ,rst four eigenvalues for various rN values (R "0)3, R "0)1, R "1, R "3)
? @ A B
rN j j j j
   
0)001 1)8287 2)8923 4)5185 4)8421
0)01 1)8280 2)8932 4)5156 4)8469
0)1 1)8215 2)9002 4)4882 4)8933
0)2 1)8143 2)9032 4)4615 4)9411
0)3 1)8073 2)9018 4)4378 4)9857
0)4 1)8005 2)8968 4)4166 5)0272
0)5 1)7941 2)8887 4)3976 5)0658
0)8 1)7765 2)8526 4)3514 5)1656
1 1)7664 2)8233 4)3280

be very carefully chosen so that the resulting model is not unduly complicated, yet is
su$ciently accurate to capture all important characteristics of the system.

7. SUMMARY AND CONCLUSIONS

A study has been made of the eigenfrequencies of a distributed rotor}shaft}beam system.

The fully coupled, non-linear governing equations are derived for a general distributed
parameter model which considers the contributions of rotor, hub, and payload inertias.
Additionally, the contribution of geometric sti!ening (softening) due to centripetal and
Coriolis accelerations are included. It is shown that variations in the ratios of system beam
inertia to rotor inertia, and beam sti!ness to shaft sti!ness, can signi"cantly a!ect
the system fundamental frequency and the associated mode shapes. Moreover, it is
 FLEXIBLE ARM ON SHAFT 697
demonstrated that a model considering the distributed nature of both the shaft and beam
are necessary if one wishes to accurately predict the system eigenfrequencies. Finally, the
a!ects of hub, shaft, and payload inertia, as well as hub rotation rate may contribute
signi"cantly to system behavior, and thus must be considered in the modelling.

REFERENCES

1. E. BARBIERI and U. OZGUNER 1988 Journal of Dynamic Systems, Measurement, and Control 110,
416}421. Unconstrained and constrained mode expansions for a #exible slewing beam.
2. A. YIGIT, A. SCOTT and A. G. ULSOY 1988 Journal of Sound and <ibrations 121, 201}210. Flexible
motion of a radially rotating beam attached to a rigid body.
3. F. BELLEZZA, L. LANARI and G. ULIVI 1990 Proceedings of IEEE Conference on Robot
Automation. 734}739. Exact modeling of the #exible slewing link.
4. A. A. SHABANA 1996 Journal of Sound and <ibration 192, 389}398. Resonance conditions and
deformable body coordinate systems.
5. K. H. LOW 1990 Journal of <ibration and Acoustics 112, 497}500. Eigen-analysis of a tip loaded
beam attracted to a rotating joint.
6. K. H. LOW 1997 Journal of Sound and <ibration 204, 823}828. A note on the e!ect of hub inertia
and payload on the vibration of a #exible slewing link.
7. E. GARCIA and D. INMAN 1991 American Institute of Aeronautics and Astronautics Journal of
Guidance 14, 736}743. Modeling of the slewing control of a #exible structure.
8. F. XI and R. G. FENTON 1994 International Journal of Robotics Research 13, 443}453. Coupling
e!ect of a #exible link and a #exible joint.
9. F. XI, R. G. FENTON and B. TABARROK 1994 Journal of Dynamic Systems, Measurement, and
Control 116, 826}831. Coupling e!ects in a manipulator with both a #exible link and joint.
10. V. POTKONJAK 1988 Robotica 6, 63}69. Contribution to the dynamics and control of robots
having elastic transmissions.
11. K. A. MORRIS and K. J. TAYLOR 1996 SIAM Review 38, 294}305. A variational calculus approach
to the modelling of #exible manipulators.
12. A. A. SHABANA 1997 <ibration of Discrete and Continuous Systems. New York: Springer: "rst
edition.
13. N. C. PERKINS and C. D. MOTE Jr 1986 Journal of Sound and <ibration 106, 451}463. Comments
on curve veering in eigenvalue problems.

APPENDIX A: TORSIONAL VIBRATION SOLUTION

The equation of motion of the shaft in Figure 1 while the system vibrates freely is [12]
uK "cJ  u , (A.1)
where u is the rotating angle of any cross-section with respect to its free end, and the over
dot symbol ( . ) and prime (  ) imply the derivations with respect to time t and position
co-ordinate x respectively. Furthermore,
G
cJ " (A.2)
o .
Q
The associated boundary conditions (BC's) are
GI u (0, t)"0, GI u (¸ ,t)#J uK (¸ , t)"0. (A.3, 4)
Q Q Q " Q
Via separation of variables, a solution to equation (A.1) in the form u(x, t)"U(x) ¹(t) is
sought. It has been shown [12] that the functions U(x) and ¹(t) are
u u
U (x)"A cos x#B sin x, ¹(t)"C cos ut#D sin ut , (A.5, 6)
cJ cJ
698 O. KOPMAZ AND K. S. ANDERSON

with u being the circular frequency of free vibration. Boundary condition (A.3) yields B"0.
Substituting this result along with equations (A.5) and (A.6) into equation (A.4) yields the
frequency equation of the system which is given by the relation

 
u u u
GI !A sin ¸ !J u A cos ¸ "0,
Q cJ cJ Q " cJ Q

or after rearranging,
J
tan g"! " g, (A.7)
J
Q
where j and J are de"ned as
Q
u
g" ¸ , J "o I ¸ . (A.8, 9)
cJ Q Q Q Q Q

Thus, equation (A.7) has countable in"nite roots g , g ,... .

 
Now, let us consider the special case where the shaft is represented by an e!ective lumped
linear spring and the hub, beam, and payload are roughly approximated by a lumped rigid
disk. Here, k represents the equivalent torsional spring coe$cient and is calculated as
R
GI
k " Q. (A.10)
R ¸
Q
The dish with the inertia J is located at one end of the spring and the equivalent inertia of
"
the shafts J at the other end as it is the case in the models of many authors. Then, the
Q
natural frequency of this system is


J #J
uJ " Q " k , (A.11)
JJ R
Q "
where the tilde over u indicates the discrete model approximation. Equation (A.11) can be
transformed into the form
u
¸ "gJ "((1#k) , (A.12)
cJ Q

where gJ again is an approximation to the exact g, if we de"ne the ratio of mass moment of
inertia as
J
k" Q . (A.13)
J
"

APPENDIX B: CHARACTERISTIC EQUATION

The shaft performs torsional vibrations. As can be easily found from standard references
(for example, reference [12]) its equation of motion is

uK "cJ  u, (B.1)

 FLEXIBLE ARM ON SHAFT 699
The associated BCs are

GI u (0, t)"!¹ #J uK (0, t), GI u (¸ , t)"!¹ , (B.2, 3)

Q B 0 Q Q P
where J , ¹ and ¹ are the rotor inertia, driving torque supplied by the motor, and
0 B P
resisting torque acting of the shaft due to the hub, respectively. The BC given by equation
(B.3) can be rewritten in a more clear form by using the equation of motion of the hub. So
applying Euler's equations to the hub yields

J uK (¸ , t)!¹ !M(0, t)#r< (0, t)"0, (B.4)

& Q P
where J , M and < are the hub rotational inertia, and the applied bending moment and
&
applied shear forces (the last two represent action on the hub by the root of the beam)
respectively. Combining equations (B.3) and (B.4) yields

GI u (¸ , t)"!J uK (¸ , t)#M(0, t)!r<(0, t). (B.5)

Q Q & Q
Now, let us con"ne our attention to a hub}beam}payload system as shown in Figure 2,
where the parameters ¸ , A , I , o , E, r, M and J all retain their meaning from section
. .
4 of this paper. A dextral set of mutually perpendicular basis vectors bK , bK , bK , are "xed in
  
the hub such that bK is parallel to the axis of rotation of the hub, bK is parallel to the
 
centerline of the beam in its undeformed position, and bK "bK ;bK . A di!erential element of
  
width ds and mass

dm"o A ds (B.6)

is located at a distance s along the beam from its root by the position vector

rBK"xbK #ybK . (B.7)

 
The acceleration of this di!erential element with respect to the Newtonian reference frame
N is

,aBK",aM# aBK#,a ;rBK#,x ;(,x ;rBK)#2, x ; mBK

"[xK !yuK !uR  (r#x)!2yR uR ] bK #[uK (r#x)#yK !uR y#2xR uR ] bK , (B.8)

 
where uR and uK are the angular velocity and angular acceleration, respectively, of the hub.
The free-body diagram for this di!erential element is shown in Figure (B.1) where M, <,
¹ and p represent the moment, shear load, tension in the beam, and the transverse applied
loads, which are, respectively, acting on this element. The unit vectors tK and nL are local
tangential and normal (transverse) directions associated with the centerline of the beam at
this di!erential element. If we treat this as an Euler}Bernoulli beam, then summing the
forces acting on this di!erential element in the nL directions yields



*< *¹ *h
F "p(s) ds! ds!¹ sin h# ¹# ds sin h# ds
L *s *s *s



*< *h *y
:p(s) ds! ds#¹ ds, where h+
*s *s *s


* *y
:p(s) ds! (M) ds#¹ ds. (B.9)
*s *s
700 O. KOPMAZ AND K. S. ANDERSON

Figure B1. Free-body diagram of di!erential beam element.

The tension in the beam ¹ is given by


¸
¹: ,aBK ) b) (o A dx)#M ,a. ) b)
x
 . 

V
¸
"o A (xK !yuK !u  (r#x)!2yR uR ) dx#M [xK !yuK !uR  (r#x)!2yR uR ] "
. *

  V 
¸ ¸
:o A uK (t) y(x, t) dx#2uR (t) yR (x, t) dx#uR (t) [ (¸!x)#r(¸ !x)]

V
#M [uK (t) y(¸ , t)#uR  (t) (r#¸ )#2uR (t) yR (¸ , t)]. (B.10)
.
If we con"ne our attention to small de#ections of an Euler}Bernoulli beam, then xR and
xK appearing in equation (B.10) may be neglected, also *y/*x+*y/*s"sin h, cos h+1, and
*</*s"M"EI *y/*s. Substituting these simpli"cations along with equations
(B.7)}(B.10) into Newton's second law of motion, for the acceleration of the di!erential
element in the nL direction, yields

F "(o A ds) (,aBK ) nL )

L
or

  V 

¸ ¸
p (s)!(EI y)# o A uK y dx#2uR yR dx#uR  [ (¸!x)#r(¸ !x)]
 y
V

#M [uK y(¸ , t)#uR  (r#¸ )#2uR yR (¸ , t)]

.
"(o A ) ([yuK #uR  (r#x)#2yR uR ] y#[uK (r#x)#yK !uR y]), (B.11)
 FLEXIBLE ARM ON SHAFT 701
subject to the boundary conditions

y(0, t)"0, y(0, t)"0, (B.12, 13)

EI y (¸ , t)"!J [uK #yK  (¸ , t)], (B.14)

.
EI y (¸ , t)"M [uK (r#¸ )#yK (¸ , t)!uR y(¸ , t)]. (B.15)
.
The "rst two BCs shown are of geometric character while the last two are the natural BCs.
Equations (B.1), (B.5), and (B.11), subject to the BCs (B.2), (B.12)}(B.15) form a coupled,
highly non-linear set of di!erential equations. As such they pose no fundamental problems
for purely numerical treatment. However, if one wishes to obtain estimates of the
eigenvalues and eigenfunctions, which may in turn be used to improve the performance of
a numerical treatment, then some simpli"cations are in order. Under the assumption of
&&small'' de#ections and &&modest'' hub rotations rates, we may reasonably neglect the yR , uR , y,
and y terms appearing in equation (B.11), yielding

EI y'4 (x, t)#o A yK (x, t)"p(x, t)!o A (r#x) uK (¸ , t), (B.16)

Q
subject to the boundary conditions

y(0, t)"0, y(0, t)"0, (B.17, 18)

EI y (¸ , t)"!J [uK #yK  (¸ , t)], (B.19)

.
EI y (¸ , t)"M [uK (r#¸ )#yK (¸ , t)]. (B.20)
.
Furthermore, if there is no forcing term p(x, t), then the change of variable

yJ (x, t)"y(x, t)#(r#x) u(¸ , t) (B.21)

Q
puts the equation of motion of the beam into a homogeneous form,

EI yJ '4 (x, t)#o A yJ (x, t)"0, (B.22)

where yJ physically represents the length of the arc tracked by any cross-section of the beam.
Solving equation (B.21) for y and substituting its corresponding expression in terms of
yJ into equations (B.17)}(B.21) yields the equation of motion and associated BC's:

yJ (0, t)"ru(¸ , t), yJ (0, t)"u(¸ , t), (B.23, 24)

Q Q
EI yJ  (¸ , t)"!J yJ G  (¸ , t) (B. 25)
.
and

EI yJ  (¸ , t)"M yJ G (¸ , t). (B.26)

.
Similarly, from the Euler}Bernoulli bending hypothesis one can write the relationships

M(0, t)"EI y (0, t)"EI yJ  (0, t), (B.27)

<(0, t)"EI y (0, t)"EI yJ  (0, t). (B.28)

Substituting equations (B.27) and (B.28) into equation (B.5) gives

GI u (¸ , t)"!J uK (¸ , t)#EI yJ (0, t)!rEI yJ (0, t). (B.29)

Q Q & Q
702 O. KOPMAZ AND K. S. ANDERSON

If one applies the method of separation of variables to equations (B.1) and (B.22), which
assumes the existence of solutions in the form of
u (x, t)"U(x) ¹(t), yJ (x, t)">(x) ¹(t), (B.30, 31)
then after some tedious manipulations one obtains that the function U(x) and >(x) must be
u u
U(x)"A cos x#B sin x, (B.32)
cJ cJ
j j j j
>(x)"a cosh x#b cos x#c sinh x#d sin x (B.33)
¸ ¸ ¸ ¸
and

  
o A 
j" u ¸ . (B.34)
EI
Note that the x variables in U(x) and >(x) are di!erent, however the function ¹(t)
appearing in equations (B.30) and (B.31) must be same in each case because it represents the
natural vibrations of the whole system, and has the form
¹(t)"C cosut#D sin ut, (B.35)
where u represents the natural frequency of the system. After substituting equations (B.32),
(B.33) and (B.35), into equations (B.2), (B.3), (B.23)}(B.26) and performing the necessary
manipulations one "nds the following relationships:


u
J u A#GI B"0, (B.36)
0 Q cJ

   
u u u u u u
GI ! sin ¸ A# cos ¸ B "J u cos ¸ A#sin ¸ B
Q cJ cJ Q cJ cJ Q & cJ Q cJ Q


 
j  j
#EI a!b#r (!c#d) ,
¸ ¸
(B.37)


u u
a#b!r A cos ¸ A#B sin ¸ "0, (B.38)
cJ Q cJ Q


u u
j(c#d)!¸ A cos ¸ #B sin ¸ "0, (B.39)
cJ Q cJ Q


j
EI (a cosh j!b cos j#c sinh j!d sin j)
¸
"J u (a sinh j!b sin j#c cosh j ) c#d cos j), (B.40)
.


j 
EI (a sinh j#b sin j#c cosh j!d cos j)
¸
"!M u (a cosh j#b cos j#c sinh j#d sin j). (B.41)
.
 FLEXIBLE ARM ON SHAFT 703
Using the parameters given by equations (1)}(8) in the text, then equations (B.36)}(B.41)
become

ajA#(bjB"0, (B.42)

¸ [j (b sin((bj)#jc cos ((bj)] A#¸ [!j (b cos ((bj)#jc sin ((bj)]B

#R ja!R j b!rN R jc#rN R jd"0, (B.43)

@ @ @ @
a#b!r cos ((bj) A!r sin ((bj) B"0, (B.44)

j c#jd!¸ cos ((bj) A!¸ sin ((bj) B"0, (B.45)

[cosh j!jR sinh j] a#[!cos j!jR sin j] b#[sinh j!jR cosh j] c

C C C
#[!sin j#j R cos j] d"0, (B.46)
C
[sinh j!jR cosh j] a#[sin j#j R cos j] b#[cosh j!jR sinh j] c
D D D
#[!cos j!j R sin j] d"0. (B.47)
D
Equations (B.42)}(B.47) produce a set of six homogeneous equations which is linear in the
six unknowns A, B, a, b, c, d. Because this is a homogeneous set of equations, the
determinant of the coe$cients' matrix must be zero in order that this system of equations
have non-trivial solutions. Hence, the j values, which make the following determinant zero
are the eigenvalues (associated with the eigenfrequencies) of the system:

g 0 0 0 0
g
R j !R j !rN R j rN R j
g @ @ @ @
g 1 1 0 0
g
g 0 0 j j
g (cosh j!jR sinh j) (!cos j#jR sin j) (sinh j!jR cosh j) (!sin j!jR cos j)
g C C C C
g (sinh j#jR cosh j) (sin j#jR cos j) (cosh j#jR sinh j) (!cos j#jR sin j)
D D D D
aj (bj g
g
¸ [j (b sin ((bj)#jc cos ((bj)] ¸ [!j (b cos ((bj)#jc sin ((bj)] g
!r cos ((bj) !r sin ((bj) g
g "0.
!¸ cos ((bj) !¸ sin ((bj) g
0 0 g
g
0 0 g

(B.48)

It can be easily seen that the determinant will be zero for j"0. This is expected because the
system is semi-de"nite. Equation (B.48) is the characteristic equation for the system whose
roots are the eigenvalues (or natural frequencies) of the system. Its expansion is not given
here due to its complexity. However, the determinant given by equation (B.48) can be
expanded by using any symbolic mathematical code (like Maple or Mathematica), as has
been done in this study.
704 O. KOPMAZ AND K. S. ANDERSON

APPENDIX C: NOMENCLATURE

For this investigation, the symbols which are important to the development of system equations of
motion and associated boundary conditions are
A cross-sectional area of beam (manipulator arm)
A cross-sectional area of shaft
1
bK , bK , bK dextral orthogonal basis vectors "xed in hub
  
cJ wave propagation velocity of torsional wave in shaft
E Young's modulus of beam material
G shear modulus of shaft (e!ective driveline) material
I area moment of inertia of beam
I area moment of inertia of shaft
Q
J rigid-body mass moment of inertia of beam about its root end
J mass moment of inertia for lumped disk crudely representing rotor}beam}payload
"
system
J mass moment of inertia of the hub which connects the beam to the shaft
&
J e!ective mass moment of inertia of payload (end e!ector#work piece)
.
J e!ective mass moment of inertia of rotor (motor armature, etc.)
0
J rigid-body mass moment of inertia of the shaft
Q
¸ length of beam
¸ e!ective length of shaft
Q
M bending moment acting on di!erential element of beam
M total mass of beam
M mass of payload (end e!ector#work piece)
.
nL unit vector normal to centerline of beam at di!erential element, lying in plane of
motion
p transverse distributed lad applied to beam
r radius of hub
rN dimensionless length
R rotational inertia ratio of beam to rotor
?
R sti!ness ratio of beam to shaft
@
R rotational inertia ratio between beam and shaft
A
R rotational inertia ratio of beam to hub
B
R rotational inertia ratio between payload and beam
C
R mass ratio of payload to the beam
D
s arc length along beam locating di!erential element
t time
tK unit vector parallel to centerline of beam of di!erential element
¹ tension in beam
¹ driving torque supplied by motor
B
¹ resisting torque felt by shaft
P
< transverse shear acting on di!erential element of beam
x local radial co-ordinate associated with location as de"ned by s
y local transverse co-ordinate associated with location as de"ned by s
yI local transverse co-ordinate which implicitly considers shaft twist.
a scaled ratio of R to R
@ ?
b scaled ratio of R to R
@ A
c scaled ratio of R to R
@ B
o mass density of beam material
o mass density of shaft
Q
j system eigenvalue

rotation of shaft
h local rotation of di!erential element of beam due to deformation of beam
u natural circular frequency of system
g beam eigenvalue