You are on page 1of 23

CHAPTER 4

Influence of Bending Stiffness


In the problem formulation for the taut string in the previous chapter, the equilibrium of forces at the damper attachment point required a discontinuity in slope at that point. Evidently, a real stay cable has a nonzero value of bending stiffness and cannot kink in this way at the damper, but must have some finite curvature at the damper attachment point. Tabatabai and Mehrabi (2000) considered the combined

effects of bending stiffness and sag on first mode damping ratios using a complex eigenvalue analysis of a cable discretized into multiple elements. They also used a database of stay cable properties from actual bridges to evaluate the range of relevant parameters, and their investigation indicated that for a significant number of stay cables the influence of bending stiffness could be significant, especially in the case of a damper located near the end of the cable. This chapter presents a detailed investigation of the influence of bending stiffness on the vibrations of a cable with attached damper by modeling the cable as a tensioned beam and developing an analytical solution to the freevibration problem.

4.1 Dynamic Stiffness Formulation


The formulation in this chapter is developed using the dynamic stiffness method, which has the following advantages for this problem:

For the important case of fixed-fixed boundary conditions, the global dynamic stiffness matrix takes on a very simple form, allowing much of

91

the solution to be carried out analytically and enabling important insights into the solution characteristics.

The formulation in terms of end displacements and slopes allows more complex boundary conditions to be readily incorporated (e.g., torsional springs at the end supports).

It can be readily extended to more complex systems (e.g., a tensioned beam with dampers and springs attached at multiple locations, or a system of parallel tensioned beams interconnected with springs and/or dampers) by simply summing the contributions from each cable segment into the global stiffness matrix.

The formulation is numerically well-conditioned and generally applicable, avoiding some difficulties that would be encountered in the higher modes by a transfer matrix formulation [To avoid such difficulties in the transfer matrix formulation Franklin (1989) used different solution technique for higher modes and lower modes].

Clough and Penzien (1975) presented the derivation of the dynamic flexuralstiffness matrix for a beam, and discussed the extension of this formulation to include the effects of axial force on transverse-bending stiffness. The formulation was developed for harmonic (non-decaying) oscillations, and hyperbolic functions are used to express the spatial variation of the solution. Leung (1985) extended the dynamic stiffness

formulation to exponentially varying harmonic oscillations for both damped and undamped beam structures (without axial force), also using hyperbolic functions for the spatial solution. Gradin and Chen (1995) used a substructuring approach, combining the

92

dynamic stiffness formulation of Leung (1985) with a transfer matrix formulation to obtain a reduced-order global dynamic stiffness matrix for beam structures (without axial force) with attached springs, masses, and dashpots. The dynamic stiffness formulation developed herein includes the effect of axial force and is developed for exponentially decaying harmonic solutions, as are expected in the case of attached dampers. Rather than using hyperbolic functions for the spatial variation of the solution, which have been observed to result in numerically illconditioned problems, the present formulation expresses the solution in terms of exponentials decaying from each end of the cable segment. The dynamic stiffness matrix will be developed for a segment of a tensioned beam with length , mass per unit length m, bending stiffness EI, and an axial tension of T as

depicted in Figure 4.1.

1
T a1 y(x) x

2
m, EI a2 T

Figure 4.1: Tensioned Beam Segment Assuming small deflections in a single plane and neglecting internal damping, the partial differential equation of motion for transverse bending vibrations is given by:
4 y 2 y 2 y EI 4 T 2 + m 2 = 0 x x t

(4.1)

93

where y(x ,t) is the transverse deflection and x is the coordinate along the cable chord axis. To solve this governing differential equation subject to the boundary conditions and the continuity and equilibrium conditions, a separable solution is assumed of the form:
y ( x, t ) = Y ( x)ei t

(4.2)

in which i = 1 and the frequency is complex in general. Substituting the assumed form of solution (4.2) into the equation of motion (4.1) yields the following ordinary differential equation in the spatial coordinate:
EI d 4Y d 2Y T + 2 mY = 0 dx 4 dx 2

(4.3)

Assuming a solution of the form


Y ( x) = Yo e x

(4.4)

yields the following equation


EI 4 T 2 + m 2 = 0

(4.5)

Eq. (4.5) is a quadratic equation in 2 and its solutions are given by


T = 2 EI
2

T 2 EI

m 2 EI

(4.6)

Because is complex in general, the argument of the square root in (4.6) is also complex, and

[ ]

represents either of the two values of the multi-valued function

( )1/ 2 . Eq. (4.6) yields four distinct values of : 1 = p , 2 = p , 3 = iq , and

4 = iq , where
T p= + 2 EI T 2 EI
2

m 2 EI

(4.7)

94

and
q= T + 2 EI T 2 EI
2

m 2 EI

(4.8)

The following identity is readily verified from (4.7) and (4.8):


p2 q2 = T EI

(4.9)

The general solution to (4.3) can then be expressed as


Y ( x) = Ae px + Be px + C eiqx + De iqx

(4.10)

Previous investigations formulated the dynamic stiffness matrix by expressing the spatial variation of the solution (4.10) using hyperbolic and trigonometric functions. However, as has been demonstrated in a previous study (Franklin 1989) and confirmed in the context of the present investigation, expressing the solution in terms of hyperbolic functions leads to a numerically ill-conditioned problem. For moderate values of their arguments, the hyperbolic sine and cosine terms take on very large values, while their difference is quite small, which results in numerical difficulties for many practical problems. It is also observed qualitatively that for large values of axial tension, the

hyperbolic terms contribute primarily in a small boundary layer region near the ends of the cable, where they allow the solution to satisfy the boundary conditions on displacement, slope, moment, and shear. For this reason, it is preferable to express the solution in the following equivalent form, with an exponential term decaying from each end of the cable segment.
Y ( x) = Ae px + Be p (
x)

+ C sin(qx) + D cos(qx)

(4.11)

95

In formulating the dynamic stiffness matrix, the spatial variation of the solution can be expressed in terms of displacements and slopes at the ends of the member using displacement shape functions:
Y ( x) = 1 y1 ( x) + 1 y1 ( x) + 2 y 2 ( x) + 2 y2 ( x)

(4.12)

The displacements and slopes at the ends of the beam segment can be related to the solution coefficients in (4.11) as follows:

1 Y (0) 1 Y (0) p = 1 = 2 Y( ) e p 2 Y ( ) pe p

e p pe p 1 p

1 0 cos ql q sin ql

0 q sin ql q cos ql

A B C D

(4.13)

This relation, which can be expressed as

= WA , can then be inverted to solve for the

solution coefficients in terms of the end displacements, A = W 1 : a1 A b B = 1 c1 C d1 D a2 b2 c2 d2 a3 b3 c3 d3 a4 b4 c4 d4

1 1 2 2

(4.14)

Explicit expressions for the terms in the matrix W-1 are given in the Appendix, and using these terms, the displacement shape functions can then be expressed as
y1 ( x) y1 ( x) y 2 ( x) y 2 ( x) = a1 a2 a3 a4 b1 b2 b3 b4 c1 c2 c3 c4 d1 d2 d3 d4 e px e p (
x)

cos qx sin qx

(4.15)

Using these shape functions, the dynamic stiffness matrix can then be formulated by enforcing equilibrium of shear force and bending moment at the ends of the beam

96

segment. Using the moment-curvature relation for a tensioned beam with the assumed separable form of solution (4.2), moment equilibrium at the two ends can be expressed as: M ( x = 0, t ) = M 1e
i t

d 2Y = EI 2 dx d 2Y = EI 2 dx

ei t
x =0

(4.16)

M ( x = , t ) = M 2e

i t

ei t
x=

(4.17)

Similarly, shear equilibrium at the two ends can be expressed as: V ( x = 0, t ) = V1ei t = EI d 3Y dx3 d 3Y dx 3 ei t
x =0

(4.18)

V ( x = , t ) = V2 ei t = EI

ei t
x=

(4.19)

Substituting into (4.16) (4.19) the expression (4.12) for Y(x) in terms of displacement shape functions and end displacements, writing the result in matrix form, and canceling the exponential terms from both sides yields the following equation:
V1 M1 V2 M2

1 (0) y (0) y (0) y (0) 1 y = EI ( ) y ( ) y ( ) y ( ) 2 y ( ) ( ) ( ) ( ) 2 y y y y


1 1 2 2 1 1 2 2 1 1 2 2

y (0) 1

y (0) 1

2 (0) y

y (0) 2

(4.20)

This dynamic stiffness relation can be expressed as F = K local , and using (4.15), the local dynamic stiffness matrix can be expressed as: p3 p2 = EI 3 p pe p 2e p p 3e p p 2e p p3 p2 0 q2 q3 sin q q 2 cos q q3 0 q 3 cos q q 2 sin q a1 b1 c1 d1 a2 b2 c2 d2 a3 b3 c3 d3 a4 b4 c4 d4 (4.21)

K local

97

The following symmetry properties of the displacement shape functions are observed:
1 (0) = y 2 ( ) y 1 ( ) = y 2 (0) y 2 ( ) y (0) = y 1 2 (0) y ( ) = y 1

(4.22) (4.23) (4.24) (4.25)

As a consequence of these properties, the terms in the third and fourth rows of the local stiffness matrix can be expressed using the terms in the first two rows: K11 K local = K 21 K13 K 23 K12 K 22 K14 K 24 K13 K 23 K11 K 21 K14 K 24 K12 K 22 (4.26)

Explicit expressions for these terms are given in the Appendix.

4.2 Fixed-Fixed Tensioned Beam with Damper


The dynamic stiffness formulation is now applied to the particular problem depicted in Figure 4.2: an axially loaded beam with fixed supports at both ends and a damper attached at an intermediate point, dividing the beam into two segments, where
2

>

. This problem is of particular interest in the context of stay cable vibration

suppression in bridges.

98

m, EI c

x1
1

x2
2

Figure 4.2: Fixed-Fixed Tensioned Beam with Damper A formulation of the problem with more general support conditions is presented in Section 4.7, but the fixed-fixed case is investigated in detail here because the problem is of smaller order, allowing a more concise presentation of the problem, and revealing many of the important features. The dynamic stiffness method is formulated in terms of displacements and slopes at the ends of each segment, and because the both the displacement and slope are constrained to be zero at the fixed supports at each end, the problem depicted in Figure 4.2 can be formulated in terms of only two unknowns: the amplitude and slope at the damper, denoted A and , respectively. The force in the damper is linearly proportional to the velocity of the beam at the damper attachment point, and can be expressed as a function of the amplitude at the damper:
Fd (t ) = c y t = ci Aei t
x=
1

(4.27)

Assembling the contributions from the two beam segments into a global stiffness matrix then yields the following equation: ( K 33(1) + K11(2) ) ( K 34 (1) + K12 (2) ) ( K 43(1) + K 21(2) ) ( K 44 (1) + K 22 (2) ) 99 A = ci A 0 (4.28)

in which the superscript indicates the number of the beam segment corresponding to each term in the global stiffness matrix. Noting the symmetry properties given in (4.26), the contributions from beam segment 1 can be expressed using the terms in the first two rows of the local stiffness matrix. Because the damper force is proportional to the amplitude at the damper, it can be moved to the left hand side of the equation, and the complex eigenvalue problem for free vibrations of the beam-damper system can be written as: ( K11(1) + K11(2) + ci ) ( K12 (2) K12 (1) ) ( K 21(2) K 21(1) ) ( K 22 (1) + K 22 (2) ) A =0 (4.29)

The complex eigenvalues then correspond to values of for which the determinant of the 2-by-2 matrix in (4.29) equals zero. Setting this determinant equal to zero and solving for the damper coefficient c yields the following equation: i ( K11(1) + K11(2) ) ( K 21(2) K 21(1) )( K12 (2) K12 (1) ) =c ( K 22 (1) + K 22 (2) ) (4.30)

Noting that the damper coefficient c is purely real, the left-hand side of (4.30) must also be purely real, which yields the following equation, independent of c.
Re 1 ( K11(1) + K11(2) ) ( K 21(2) K 21(1) )( K12 (2) K12 (1) ) ( K 22 (1) + K 22 (2) ) =0

(4.31)

Eq. (4.31) will be referred to as the phase equation, and is analogous to the equation referred to by the same name in the treatment of the vibrations of a taut string with attached damper in Chapter 3. Solution branches to (4.31) give permissible values of the complex frequency for a given
1

/ L , thus revealing the attainable values of modal

damping with their corresponding oscillation frequencies, and the evolution of these solution branches under varying parameters is helpful in characterizing the response of the system. It is important to note that the contributions to (4.30) and (4.31) from the 100

local stiffness matrix for each beam segment (e.g., K11(1) and K11(2) ) are functions of p and q, and consequently, they also depend on the complex frequency . The real and imaginary parts of the complex frequency will be denoted as follows

= Re( ) = Im( )

(4.32) (4.33)

These quantities are analogous to the previous definitions for the taut string in (3.8a,b):

is the oscillation frequency and is the rate of exponential decay. However, positive
values of correspond to decaying oscillation with the presently assumed form of solution in (4.2), whereas negative values of corresponded to decaying oscillation for the taut string. Also, (4.32) and (4.33) are dimensional, in contrast with the

nondimensional definitions in (3.8a,b). Alternative nondimensional versions of (4.32) and (4.33) will be introduced in subsequent sections.

4.2.1 Nondimensionalization of Stiffness Matrix


To achieve results of general applicability and to facilitate presentation, it is useful to nondimensionalize the problem formulation. The amplitude at the damper A is nondimensionalized by the cable length:
A = ( A / L)

(4.34)

The terms in (4.29) from the stiffness matrix for each segment are normalized by the bending stiffness and the length:
(k ) (k ) k 11 = ( L3 / EI ) K11 (k ) (k ) k 12 = ( L2 / EI ) K12

(4.35) (4.36)

101

(k ) (k ) k 21 = ( L2 / EI ) K 21 (k ) (k ) k 22 = ( L / EI ) K 22

(4.37) (4.38)

With these normalizations, the complex eigenvalue problem (4.29) can then be rewritten in nondimensional form as: [k11(1) + k11(2) + i (cL3 / EI ) ] (k12 (2) k12 (1) ) (k21(2) k21(1) ) (k22 (1) + k22 (2) ) A =0 (4.39)

Explicit expressions for the nondimensional terms in the stiffness matrices (4.35) (4.38) are given in the Appendix; these terms are functions of nondimensional versions of the mode shape parameters p (4.7) and q (4.8):
p = pL q = qL

(4.40) (4.41)

Alternative explicit expressions for p and q are given in the following sections, depending on the choice of nondimensionalization for the complex frequency . Nondimensionalization of the identity in (4.9) yields the following relation: p2 q2 = 2 (4.42)

where is a nondimensional bending stiffness parameter, as used by Tabatabai and Mehrabi (2000):

TL2 EI

(4.43)

When becomes large, bending effects are less significant, and the tensioned beam behaves more like a taut string; when is small, bending effects predominate. Using a database of stay-cable properties, Tabatabai and Mehrabi (2000) report that nearly all

102

bending stiffness parameters ( ) are within the range of 10-600, with 82% of the cables having values of larger than 100. The appropriate choice of nondimensionalization for the complex frequency depends on the magnitude of . In most cases a taut-string nondimensionalization will be used, to facilitate comparison with the taut-string results, but in cases of zero tension, the taut-string nondimensionalization cannot be used, and an alternative beam nondimensionalization will be employed.

4.2.2 Beam Nondimensionalization of Frequency


When is small, indicating that the axial tension T is small relative to the bending stiffness EI / L2 , the undamped natural frequencies are close to the natural frequencies of a fixed-fixed beam, and it is helpful to introduce the following beam nondimensionalization of the frequency:

2
L2

EI m

(4.44)

Substituting (4.44) into the definitions for p (4.7) and q (4.8), the following expressions for the nondimensional values p (4.40) and q (4.41) can be obtained:
p = 2 / 2 + ( 2 / 2) 2 + ( 2 )2 q = 2 / 2 + ( 2 / 2) 2 + ( 2 ) 2

(4.45) (4.46)

Using this nondimensionalization of frequency, the following expression for the nondimensional damper coefficient, analogous to (4.30), can be obtained:

103

(2) (1) (2) (1) (1) + k (2) ) (k21 k21 )(k12 k12 ) = ( k 11 11 (1) + k (2) ) 2 (k 22 22 i

cL EI m

(4.47)

From (4.47) the following beam nondimensional version of the phase equation (4.31) can be obtained:
Re 1 (k
(1) 11

+k

(2) 11

(k21(2) k21(1) )(k12 (2) k12 (1) ) ) (k22 (1) + k22 (2) )

=0

(4.48)

Under the beam nondimensionalization, the real and imaginary parts of the complex frequency will be denoted and , respectively, as in (4.32) and (4.33). When

0 , corresponding to a beam without axial tension, p and q are equivalent and can
be expressed as:
p = q =

(4.49)

4.2.3 Taut-String Nondimensionalization of Frequency


When is large, indicating that the tension T is large relative to the bending stiffness EI / L2 , the natural frequencies in the absence of the damper are close to the natural frequencies of a taut string, and it is convenient to introduce the following nondimensionalization of frequency: T = L m

(4.50)

From (4.50) and (4.44) it can be shown that the alternative nondimensional frequencies are related by
= ( / )

(4.51)

104

It is also noted that ( / ) is equal to the ratio of the fundamental frequency of a taut string ( / L) T / m to the fundamental frequency of a pin-supported beam Using (4.50), p (4.40) and q (4.41) can be expressed as:
)2 p = pL = 2 / 2 + ( 2 / 2)2 + ( )2 q = qL = 2 / 2 + ( 2 / 2) 2 + (

(4.52) (4.53)

Using this taut-string nondimensionalization of frequency, the following expression for the nondimensional damper coefficient, analogous to (4.30), can be obtained: i 2 (k11(1) + k11(2) ) (k21(2) k21(1) )(k12 (2) k12 (1) ) c = (1) (2) (k22 + k22 ) Tm (4.54)

From (4.54) the following taut-string nondimensional version of the phase equation (4.31) can be obtained:
Re (k (2) k21(1) )(k12 (2) k12 (1) ) 1 (k11(1) + k11(2) ) 21 (k22 (1) + k22 (2) ) =0

(4.55)

Under the taut-string nondimensionalization, the real and imaginary parts of the
will be denoted and , respectively, as in (4.32) and (4.33). complex frequency

4.2.4 Nondimensional Mode Shapes


For the fixed-fixed beam, the mode shapes can be expressed in terms of the nondimensional amplitude A and the slope at the damper location. The

nondimensional mode shape over the shorter cable segment can be expressed as
(1) (1) Y (1) ( x1 ) = A y ( x1 ) + y ( x1 ) 2 2

(4.56)

And the nondimensional mode shape over the longer segment can be expressed as

105

(2) (2) Y (2) ( x2 ) = A y ( x2 ) + y ( x2 ) 1 1

(4.57)

The shape functions in (4.56) and (4.57) are nondimensional versions of those in (4.15) and are given by
(k ) y ( xk ) 1 (k ) y ( xk ) 1 (k ) y ( xk ) 2 (k ) y ( xk ) 2 (k ) (k ) (k ) a1 b1( k ) c1 d1

e pxk / L
k

(k ) (k ) a (2k ) b 2 c2 d (2k ) e p (

xk ) / L

(k ) (k ) (k ) (k ) cos qxk / L a3 b3 c3 d3 a ( k ) b ( k ) c ( k ) d ( k ) sin qxk / L 4 4 4 4

(4.58)

Explicit expressions for each term of the coefficient matrix in (4.58) are given in the Appendix. For a given value of the complex frequency , the amplitude and slope at the damper of the corresponding mode shape at the damper can be related by the following equation, obtained from the second row of (4.39): (k21(1) k21(2) ) A = (k22 (1) + k22 (2) ) (4.59)

4.3 Non-Oscillatory Decaying Solutions


In the case of a taut cable without bending stiffness, it was previously observed in Section 3.3.1 that solutions exists for which the cable decays without oscillation. Such solutions also exist for the tensioned beam, and for these solutions, the frequency is zero and is purely imaginary: = i . Using the taut-string nondimensionalization, p (4.52) and q (4.53) can then be rewritten as:
)2 p = 2 / 2 + ( 2 / 2) 2 ( )2 q = 2 / 2 + ( 2 / 2) 2 (

(4.60) (4.61)

Similarly, (4.54) can be rewritten as

106

1 2

(k11(1) + k11(2) )

(k21(2) k21(1) )(k12 (2) k12 (1) ) c = (1) (2) (k22 + k22 ) Tm

(4.62)

For given values of and

/ L , the nondimensional damper coefficient c / Tm can be

; Figure 4.3 shows a resulting plot of computed from (4.62) over a range of values of versus c / Tm for
1

/ L = 0.05 and for several different values of the bending

stiffness parameter . The curve corresponding to a taut string with zero bending stiffness plotted previously in Figure 3.2 is also plotted with these curves for reference. The curves corresponding to = 1000 and to the taut-string result terminate

on the plot because numerical difficulties were encountered in evaluating the solution for
, not because the solution actually ceases to exist. large values of

Similar to the taut-string result, it is evident in Figure 4.3 that zero-frequency solutions only exist when c / Tm is greater than some critical value. In the taut-string case that critical value was c / Tm = 2 , and Figure 4.3 shows that this critical value is increased by the influence of bending stiffness. The critical value of c / Tm for the case of = 10 is indicated in Figure 4.3 by a vertical dotted line at the lowest value of c / Tm for which a solution exists. In contrast with the zero-frequency solution for the taut string, for which only one solution for the decay rate existed for any given
exist for each supercritical value of supercritical value of c / Tm , two solutions for

c / Tm when the bending stiffness is nonzero. A vertical dotted line is plotted in Figure 4.3 at a supercritical value of c / Tm for the case of = 10 , and the two solutions are
, which corresponds to a more quickly indicated with circles. The larger value of is denoted decaying solution, is denoted the fast solution, and the smaller value of

107

the slow solution. This behavior can be compared with that of the zero-frequency solution for the SDOF oscillator, plotted in Figure 3.3, for which two solutions for the decay rate exist for supercritical values of the damper coefficient. Figure 4.4 shows the evolution with increasing c / Tm of the mode shapes associated with the fast and slow solution branches. Both branches begin at the critical value of c / Tm with the
, so initial mode shapes, plotted with a heavier line in Figure 4.4, are same value of

identical. The evolution of the slow solution is similar to that previously observed for the taut string in Figure 3.4, approaching the static deflected shape of the beam under a concentrated load at the damper location as c / Tm .

1000

:
10 50 "fast" solution 100 1000 Taut String

100

10


1 "slow" solution 0.1

/ L = 0.05 1

0.01 1 2 10 "critical" damping 100 1000

c Tm

Figure 4.3: Nondimensional Decay Rate vs. Nondimensional Damper Coefficient with Varying Bending Stiffness ( 1 / L = 0.05 , Taut-String Nondimensionalization)

108

a) 1

"slow" solution

=100
increasing c / Tm
1

/ L = 0.3

Y ( x)

0 0 0.2 0.4 0.6 0.8 1

b) 1 "fast" solution

=100
1

/ L = 0.3

Y ( x)
increasing

c / Tm

0 0 0.2 0.4 0.6 0.8 1

x/ L

Figure 4.4: Evolution of Slow and Fast Non-Oscillatory Mode Shapes with Increasing Nondimensional Damper Coefficient ( = 100 , 1 / L = 0.3 )

Figure 4.5 is similar to Figure 4.3, but corresponds to a damper located further from the end of the cable:
1

/ L = 0.3 . The critical values of c / Tm are closer to the

taut-string result in this case than in Figure 4.3. Unlike the taut-string case, for which the critical value ccrit / Tm is independent of damper location, for a given value of the bending stiffness parameter , ccrit / Tm varies with
1

/L.

109

1000

:
10 50 100 1000

100

10

/ L = 0.3 1

Taut String

0.1

0.01 1 2 10 100 1000

c Tm
Figure 4.5: Nondimensional Decay Rate vs. Nondimensional Damper Coefficient with Varying Bending Stiffness ( 1 / L = 0.3 , Taut-String Nondimensionalization) The critical value of the nondimensional damper coefficient ccrit / Tm can be computed for given values of
1

/ L and the bending stiffness parameter by computing

, and determining the minimum value of c / Tm , as c / Tm over a range of values of

indicated schematically in Figure 4.3 for the curve corresponding to = 10 . Figure 4.6 shows a contour plot of ccrit / Tm , generated in this manner over a range in values of the damper location (from
1

/ L = 0.005 to

/ L = 0.1 ) and bending stiffness parameter

(from = 10 to = 1000 ). For a given value of , it can be seen that ccrit / Tm decreases toward the taut-string value of 2 as
1

/ L increases. This indicates that the

influence of bending stiffness is most significant when the damper is located near a fixed 110

support, as may be expected, because, as has been previously noted (e.g. Franklin 1989), the bending effects for a tensioned beam are most significant in the region near the supports.

Figure 4.6: Contour Plot of the Critical Value of the Nondimensional Damper Coefficient ccrit / Tm vs. Bending Stiffness Parameter and Damper Location In the case of a beam without axial tension ( = 0 ), using (4.49) with = i (beam nondimensionalization) p and q can written as p = q = (i + 1) 2 (4.63)

For a given damper location, the nondimensional damper coefficient cL / EI m can then be computed from (4.47) over a range of values of (evidently, the

nondimensionalization c / Tm is not appropriate when T 0 ), and as previously, the 111

critical value of the nondimensional damper coefficient is given by its minimum value. Figure 4.7 shows a plot of the critical value of the nondimensional damper coefficient ccrit L / EI m against the damper location
1

/ L for the fixed-fixed beam with zero

tension. The critical value ccrit L / EI m takes on very large values as the damper approaches a fixed support and decreases to a value of 16.619 when the damper is near midspan. Figure 4.8 shows a plot of the critical value of the nondimensional damper coefficient normalized by the damper location ccrit L / EI m ( 1 / L) against this normalized critical value is virtually constant for critical value ccrit L / EI m is inversely proportional to
1 1

/ L , and

/ L < 0.2 , indicating that the

/ L over this range.

10000

1000
ccrit L EI m

100

16.619

10 0 0.1 0.2
1

0.3

0.4

0.5

/L

Figure 4.7: Critical Value of the Nondimensional Damper Coefficient ccrit L / EI m vs. Damper Location for a Beam with Zero Tension ( = 0 ) 112

8.4 8.2 8 7.8


ccrit L 1 EI m L

7.6 7.4 7.2

7 6.934 6.8 0 0.1 0.2


1

0.3

0.4

0.5

/L

Figure 4.8: Normalized Critical Value of the Nondimensional Damper Coefficient ccrit L / EI m ( 1 / L) vs. Damper Location for a Beam with Zero Tension ( = 0 )

4.4 Limiting Cases of Non-Decaying Oscillation


In the case of a taut cable without bending stiffness, it was previously observed that there are solutions for which the cable oscillates without decay; these solutions were associated with the limits of c 0 and c . For such solutions, the frequency is purely real, = .

4.4.1 Undamped Modes


When c 0 , the problem reduces to computing the natural frequencies and mode shapes of a fixed-fixed tensioned beam. This problem has been previously solved and is discussed by Wittrick (1986), who presents an iterative solution technique for the 113

You might also like