You are on page 1of 12

Smart Materials and Structures

Related content
- Clipped viscous damping with negative
Energy based optimization of viscous–friction stiffness for semi-active cable damping
F Weber and C Boston
dampers on cables - Cycle energy control of
magnetorheological dampers on cables
F Weber, H Distl, G Feltrin et al.
To cite this article: F Weber and C Boston 2010 Smart Mater. Struct. 19 045025
- Optimal semi-active damping of cables:
evolutionary algorithms and closed-
formsolutions
C Boston, F Weber and L Guzzella
View the article online for updates and enhancements.

Recent citations
- Modelling and passive control of flexible
guiding hoisting system with time-varying
length
Naige Wang et al

- Wind vibration control of stay cables using


magnetorheological dampers under
optimal equivalent control algorithm
Yu-Liang Zhao et al

- Curved surface sliders with friction


damping, linear viscous damping, bow tie
friction damping, and semiactively
controlled properties
Felix Weber et al

This content was downloaded from IP address 195.50.133.83 on 10/01/2020 at 18:44


IOP PUBLISHING SMART MATERIALS AND STRUCTURES
Smart Mater. Struct. 19 (2010) 045025 (11pp) doi:10.1088/0964-1726/19/4/045025

Energy based optimization of


viscous–friction dampers on cables
F Weber1,3 and C Boston1,2
1
Swiss Federal Laboratories for Materials Testing and Research, Structural Engineering
Research Laboratory, Ueberlandstrasse 129, CH-8600 Duebendorf, Switzerland
2
Institute for Dynamic Systems and Control, ETH Zurich, 8092 Zurich, Switzerland

E-mail: felix.weber@empa.ch and charles.boston@empa.ch

Received 23 December 2009, in final form 29 January 2010


Published 19 March 2010
Online at stacks.iop.org/SMS/19/045025

Abstract
This investigation optimizes numerically a viscous–friction damper connected to a cable close
to one cable anchor for fastest reduction of the total mechanical cable energy during a free
vibration decay test. The optimization parameters are the viscous coefficient of the viscous part
and the ratio between the friction force and displacement amplitude of the friction part of the
transverse damper. Results demonstrate that an almost pure friction damper with negligibly
small viscous damping generates fastest cable energy reduction over the entire decay. The ratio
between the friction force and displacement amplitude of the optimal friction damper differs
from that derived from the energy equivalent optimal viscous damper. The reason for this is that
the nonlinearity of the friction damper causes energy spillover from the excited to higher modes
of the order of 10%, i.e. cables with attached friction dampers vibrate at several frequencies.
This explains why the energy equivalent approach does not yield the optimal friction damper.
Analysis of the simulation data demonstrates that the optimally tuned friction damper dissipates
the same energy per cycle as if each modal component of the cable were damped by its
corresponding optimal linear viscous damper.

1. Introduction brief description shows that the characteristics of commonly


used cable dampers range from almost pure linear viscous to
In recent years, stay cable vibration mitigation has become almost pure friction damping, with numerous viscous–friction
an important issue due to the increasing free span of stay combinations in between.
cable bridges. Besides elastomer bearings within the anchor The engineer’s task is to design these dampers for the
system, the most common vibration mitigation method is maximum damping ratio of the most critical cable modes for
the connection of external transverse dampers to cables a given damper position. If the damper behaviour can be
near their lower anchor. Commonly used damper types approximated by the linear viscous damper model, the optimal
are oil (Sun et al 2005), sliding surface (Bournand and tuning of the damper coefficient c for one target mode is known
Crigler 2005) and magneto-rheological (Weber et al 2005b, (Krenk 2000, Krenk and Høgsberg 2005), and optimal control
Christenson et al 2006, Li et al 2007, Neelakantan and based methods were used to find a solution to the problem
Washington 2008). Oil dampers mainly provide linear viscous for several modes (Wang et al 2005). Design rules have also
damping superposed by a relatively small Coulomb friction been derived for cables including flexural rigidity or for finite
damping due to the damper sealing. Sliding surface dampers damper support stiffness (Fujino and Hoang 2008). Even for
show predominantly Coulomb friction characteristics with a
nonlinear viscous dampers, approximate closed-form solutions
small viscous behaviour produced by lubricants between the
for the optimal damper tuning exist (Main and Jones 2002).
surfaces. Magneto-rheological dampers at constant current
If a pure friction damper is used, investigations show that the
provide a strongly current dependent hysteretic friction force
friction force has to be adjusted in proportion to the damper
that is superposed by a weakly current dependent viscous
displacement amplitude (Inaudi 1997, Guglielmino et al 2004,
force (Sims et al 2004, Weber et al 2005a, 2008). This
Boston et al 2009) in order to provide approximately amplitude
3 Author to whom any correspondence should be addressed.
independent damping to the cable and to avoid clamping (He

0964-1726/10/045025+11$30.00 1 © 2010 IOP Publishing Ltd Printed in the UK & the USA
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 1. Force displacement and force velocity trajectories of linear viscous (a), Coulomb friction (b) and viscous–friction (c) dampers.

et al 2003). Despite the amplitude proportional control of the where u̇ a is the velocity of the damper and of the cable at the
friction force, the nonlinearity of the friction damper remains, damper position and the proportional factor c is the damper’s
and thereby leads to energy spillover from the excited mode viscous coefficient (figure 1(a)). The minus sign in (2)
to higher modes (Weber et al 2009). In contrast to the two guarantees that the external damper acts as a dissipative device.
prototype dampers, i.e. linear viscous and friction dampers Please notice that this sign convention is used consistently
(Vader and McDaniel 2007), the optimal tuning of dampers in all equations and figures of this paper. Since the damper
with both linear viscous and friction components for maximum velocity is determined by the interaction between damper and
cable damping has so far not been investigated. cable, the tuning parameter of the viscous damper is its viscous
The present investigation tries to fill this gap with an coefficient c. If the goal is to maximize cable damping or
optimization study of a transverse viscous–friction damper on to maximize the energy reduction rate in the cable over a
a cable. The damper characteristics are varied from those free decay test with an external viscous damper, the viscous
of a pure friction device to those of a pure viscous device, coefficient c has to be tuned according to the following optimal
by increasing the damper’s viscous coefficient and decreasing closed-form solution (Krenk 2000)
its friction component. The optimization criterion is the
T
maximum cable energy reduction rate over a free vibration cnopt ∼
= for a/l  2% (3)
decay test. The optimal damper parameters are derived as a 2π f rn0
a function of the cable energy reduction range. A damper
cycle energy based explanation for the optimal damper tuning where f rn0 represents the frequency in hertz of the undamped
parameters is given. mode n and l is the cable length.

2. Combined viscous–friction damper on cable 2.3. Friction damper


The Coulomb friction damper, hereafter simply referred to as
First, the equation of motion of a cable with a transverse
the friction damper, generates a force that only depends on the
external damper is introduced, and then the characteristics of
sign of the damper velocity (figure 1(b))
linear viscous, friction and combined viscous–friction dampers
are described. f = −sgn(u̇ a )Ffri (4)

2.1. Cable with local transverse force where Ffri is the friction force level. Due to the nonlinearity
of the friction damper, a closed-form solution for its optimal
The equation of motion of a taut string with an external tuning for maximum cable damping is not available. For this
transverse force f is (Irvine 1981) reason, a common approach is to balance the cycle energies W
T u  − m ü = f δ(x − a) (1) of viscous and friction dampers for equal damper displacement
amplitudes Ua (Weber et al 2009)
where T denotes the cable tension force, u = u(x, t) is the
transverse cable displacement, δ is the Dirac delta function that W = πcU̇a Ua = 4 Ffri Ua (5)
applies the force to the cable at position x = a and the inherent
damping of the taut string is assumed to be zero. The prime and where U̇a denotes the damper velocity amplitude and the
the dot represent spatial and time derivatives, respectively. amplitude is understood as the peak value of a vibrating
state during one period of vibration. Replacing the viscous
2.2. Linear viscous damper coefficient in (5) by its optimum (3) and assuming sinusoidal
damper displacement u a (t) and consequently single mode
If the external force f is generated by a linear viscous damper, vibrations in the cable–damper system, the friction force
commonly only denoted as a viscous damper, then the force f becomes
becomes πT
Ffri = Ua . (6)
f = −cu̇ a (2) 4a

2
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Equation (6) shows that the force level of a friction damper, 3. Damper optimization procedure
which dissipates the same cycle energy as the optimal linear
viscous damper and is derived on the assumption of single 3.1. Free decay tests
mode vibrations in the entire cable–damper system, has The task is to optimize the local viscous–friction damper
to be adjusted in proportion to the damper displacement for maximum cable damping. Maximum cable damping is
amplitude Ua with the proportional factor (π T )/(4a). indicated by:
However, numerical and experimental investigations with
(i) minimum steady state response of the cable, or
friction dampers on cables have shown that the nonlinear
(ii) fastest decrease of vibration amplitudes during a free
nature of friction dampers introduces higher frequencies in
vibration decay test.
the cable by energy spillover from the excited mode to higher
modes (Weber et al 2009, 2010). Therefore, the assumptions The criterion (i) was tested in simulation by excitation of
of sinusoidal damper displacement and single mode vibrations the cable at its fundamental undamped resonance frequency
f r10 up to steady state conditions. The connected damper
used for the derivation of (6) are not fulfilled. This implies that
was a linear viscous type. Each simulation was performed
(π T )/(4a) does not represent the optimal proportional factor
at a different viscous coefficient c, going from a very small
of amplitude proportional friction dampers for maximum cable
value c1min ≈ 0 to a very large value c1max . It turned out that
damping. This can be compensated by the introduction of the
the steady state amplitude of the cable midspan displacement
optimization parameter β as follows opt
was not minimized at c = c1 but at c1max > c1 . The
opt

reason for this seemingly contradictory result is that the cable


πT
tune
Ffri = Ua β (7) fundamental frequency also changes from f r1min to f r1max
4a when going from c1min to c1max due to the shortening effect
in order to optimize the amplitude proportional friction of the cable by the damper (Krenk and Høgsberg 2005).
Thus, the cable is only excited precisely at its fundamental
damper for maximum cable damping. Similar proce-
frequency if the viscous damper with c1min ≈ 0 is tested. The
dures can be found in Bhaskararao and Jangid (2006),
test of viscous dampers with c > c1min would require us to
Ciǧeroǧlu and Oz̈guv̈en (2006), Ye and Williams (2006), Choi
precisely match the excitation frequency to the actual cable
et al (2008), and Weber et al (2009). resonance frequency. This problem could only be solved by
time consuming iterations. In order to avoid this, the second
2.4. Combined viscous–friction damper (ii) approach is chosen here and the cable damper system is
always excited with the same excitation forces. After turning
The force of the combined viscous–friction damper is chosen off the excitation forces, the damper working behaviour is
as follows (figure 1(c)) assessed by the free decay response of the cable.
  
π T 3.2. Cable damper model
f = −γ cu̇ a + sgn(u̇ a )Ua − c 2π f r n
0
(8)
4 a
Numbers of free vibration decay tests at different damper
where c tunes the viscous damping part only and γ optimizes settings are simulated using a linear truss element cable model
with a transverse viscous–friction damper as follows (Bathe
the sum of the viscous and friction damping parts. The
1982)
notation (8) differs from the expressions for linear viscous
Mü + Cu̇ + Ku = fw + a f (9)
damping (3) and friction damping (7):
where u denotes the vector of the cable nodal displacements
(1) The friction part is lowered by the term Ua (π/4)c2π f rn0 . and M, C and K are the mass, structural damping and stiffness
matrices, respectively. The inherent cable damping is assumed
This generates energy equivalent viscous–friction damp-
to be zero, i.e. C = 0. The mass and stiffness matrices are
ing for arbitrary values of c if γ = 1.
tri-diagonal matrices as follows
(2) The optimization parameter γ does not only tune the ⎡ ⎤
2/3m ele 1/3m ele 0 ··· 0
friction part as the parameter β the friction force in (7)
⎢ .. .. .. ⎥
but γ influences the sum of friction and viscous damping. ⎢ 1/3m ele . . . ⎥
⎢ ⎥
The reason for this is explained as follows: if γ only M=⎢ 0 ⎢ . .. ⎥
0 ⎥ (10)
tunes the friction part, then the friction part becomes zero ⎢ .. ⎥
opt ⎣ . 1/3m ⎦
for c = cn and for arbitrary values of γ . In contrast, ele
if γ multiplies the sum of viscous and friction damping 0 ··· 0 1/3m ele 2/3m ele
opt ⎡ ⎤
according to (8), c = cn will only produce optimal linear 2T /L ele −T /L ele 0 ··· 0
viscous damping if γ = 1. Since the cable damping ratio ⎢ .. .. .. ⎥
⎢ −T /L ele . . . ⎥
resulting from optimal linear viscous damping is known ⎢ ⎥
K=⎢ .. ⎥
according to Krenk (2000), the numerical results due to ⎢ 0 . 0 ⎥
⎢ .. ⎥
opt
c = cn and γ = 1 can be used later to validate the ⎣ . −T /L ele ⎦
damper optimization procedure that is used to seek for the 0 ··· 0 −T /L ele 2T /L ele
optimally tuned combined viscous–friction damper. (11)

3
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 2. Approximated friction force trajectory (a), Matlab/Simulink® model of cable with viscous–friction damper (b).

where m ele is the mass of one finite element and L ele is the friction force also jumps between its positive and negative
its length. The cable is modelled with 200 elements which values almost each time step. This leads to both excessive
correspond to 199 nodal points. The vector fw is the computing time and scattering effects of the simulated states.
disturbance force vector, a is the damper location vector and In order to avoid these drawbacks, the friction part of the
f is the damper force according to (8). The damper is located combined viscous–friction damper in (8) is approximated by
at 2% cable length, the following equation
a/l = 2%, (12) ⎧

⎪ −γ sgn(u̇ a )Ffri : |u̇ a | > u̇ a−thresh

⎨  
since this is a typical damper position and allows for the use (|u̇ a | − u̇ a−thresh )6
of (3) when tuning the viscous damping part. The cable f = −γ sgn(u̇ a )Ffri 1 − : (14)

⎪ u̇ a−thresh
properties are close to the values of a real cable and are chosen ⎪

such that the fundamental frequency f r10 is 1 Hz: the cable |u̇ a |  u̇ a−thresh
tension force is 2500 kN, the length is 100 m and the mass that applies the desired friction force to the cable for damper
per unit length is 62.5 kg m−1 . The cable is excited at its velocities larger than the threshold velocity u̇ a−thresh and
fundamental frequency f r10 by the disturbance force vector fw generates a highly nonlinear viscous force by the power
assuming a sinusoidal mode shape, thus function of sixth order for damper velocities smaller or equal
to u̇ a−thresh . The formulation of (14) guarantees that the two
1 sin(π x/l) sin(2π f r10 t): t < 140 s branches of the force velocity trajectory have the same value
fw (x, t) =
0: t  140 s and slope at u̇ a−thresh . The threshold velocity is not a constant
(13) value but is adjusted in proportion to the friction force level Ffri
where the vectors 1 and 0 include all nodal points of the finite as follows
element model along the cable axis. The disturbance force u̇ a−thresh = Ffri /2 × 106 . (15)
is turned off at 140 s in order to simulate the free vibration This adaptation guarantees that the error of the approxima-
decay test. The sinusoidal mode shape is only fully correct tion (14) does not depend on the value Ffri . This is depicted
for the case of a linear cable with linear damper. This is in figure 2(a) by the plotted force velocity trajectories and
given for the cases where the external damper behaves as a their close-ups for two different values of Ffri . The factor of
opt
linear viscous damper with c = cn and arbitrary values of γ , 2 × 106 was found by simulation. The models of the cable, the
see (8). In contrast, if the friction part of the external damper damper, the amplitude proportional friction force tuning and
is different from zero, the nonlinear behaviour of the damper the disturbance force are programmed in Matlab/Simulink®
will slightly change the mode shape. However, since the (figure 2(b)). The nonlinear system is solved by the ode45
free vibration decay tests are not evaluated by the logarithmic (Dormand-Prince) solver with a variable time step size which
decrement method, which only uses the information of one is constrained by the upper limit of 1 × 10−3 s.
nodal point of the cable, but by the decay rate of the total
mechanical cable energy, the changed mode shape hardly
3.3. Simulation of free decay tests
affects the damper optimization procedure. The friction part
in (8) could be calculated using the ‘Coulomb friction’ function Some simulated states of a free vibration decay test with a
of the Matlab/Simulink® library. However, during simulation combined viscous–friction damper with small viscous damping
opt
phases where the damper velocity is close to zero, the sign at c/c1 = 0.04 and γ = 1.2 are plotted in the next
of the velocity changes almost each time step. Consequently, figures. The midspan cable displacement increases and

4
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 3. Simulated free vibration decay test: midspan cable displacement (a), damper characteristics (b).

Figure 4. Simulated free vibration decay test: damper displacement and velocity (a), damper force components (b).

reaches steady state conditions as long as the disturbance as the damper force balances the force produced by the cable
force excites the cable (figure 3(a)). The characteristics in the damper direction. During these phases, the damper
of the external damper are depicted by the force velocity velocity and consequently the viscous damper force are zero
trajectories during the entire decay in figure 3(b). These (figure 4(b)). The working phases of the damper are indicated
trajectories show constant slope due to the viscous part of by the velocity peaks which generate the peak behaviour of the
opt
c/c1 = 0.04. The friction part can be read off from the viscous damper force.
point where the trajectories meet the y -axis. The friction
part decreases during the decay phase since it is adjusted 3.4. Assessment of free vibration decay tests
in proportion to the decreasing cable amplitude at damper
Free vibration decay tests are usually assessed by the ratios of
position Ua (t) according to (8). The damper displacement
adjacent vibration peaks at an anti-node position. The ratios
amplitude is determined using the function ‘weighted moving
yield the point-to-point logarithmic decrement (Huanga et al
average’ from the Matlab/Simulink® library. This produces
2007)  
a smooth envelope of the displacement amplitudes as can be u(ti , l/2)
seen in figure 4(a) and therefore a smooth friction force level δi = ln (16)
u(ti + Td , l/2)
(figure 4(b)). In contrast to pure linear viscous damping, which
evokes a sinusoidal displacement at the damper position, the where Td denotes the time period of the damped vibration,
friction part of the combined viscous–friction damper leads to which leads to the point-to-point damping ratios as follows
almost rectangular cable displacement at the damper position δi
(figure 4(a)). The reason for this phenomenon is that the ζi =  . (17)
friction force holds the cable in its extreme positions as long 4π 2 + δi2

5
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 5. Displacement decay at the anti-node position and corresponding point-to-point damping ratios for optimal linear viscous
damping (a) and pure friction damping (b).

Figure 6. Displacement decay at 87% cable length and the corresponding point-to-point damping ratios (a) and energy spillover (b) for pure
friction damping.

If the cable damper system under consideration is linear, on the value of γ , the damper transfers up to 10% of the total
i.e. the external damper is a pure viscous damper, the cable mechanical cable energy E tot (18) from the excited mode to
damper system vibrates at the single frequency of excitation. higher modes (figure 6(b))
Then, vibration amplitudes decay with an exponential envelope
(Weber et al 2009) and the point-to-point damping ratios E tot (t) = 12 uT (t)Ku(t) + 12 u̇T (t)Mu̇(t). (18)
are constant during the entire decay phase, as depicted in
Numerical investigations show that if the disturbance fw
figure 5(a). If the viscous damper at 2% cable length is
excites the first mode, energy is transferred to modes 3, 5, 7,
optimally tuned according to (3), the damping ratios show
etc; if mode 2 is excited, energy is transferred to modes 6, 10,
the expected value, i.e. ζ (c = cn ) ∼
opt
= a/(2l) (Krenk
14, etc; if mode 3 is excited, energy is transferred to modes
2000). Notice that the point-to-point values are determined
9, 15, 21, etc. Thus, energy spillover is the transfer of energy
until vibrations reach 2% of their amplitude at the start of
from the excited mode with wavelength λw to modes of energy
decay since simulation results become falsified below 2% due
spillover λspi with wavelengths λw /3, λw /5, λw /7 etc, which
to the approximation (14). If friction damping is present,
is expressed in (19).
the point-to-point damping ratios vary during the decay phase
and also depend on the position on the cable that is used for λspi = λw /[3, 5, 7, . . .]. (19)
their determination (figures 5(b) and 6(a)). This dependency
directly results from energy spillover due to the nonlinearity The symmetrical modes lead to a rather triangular mode
of the friction damper (Weber et al 2009, 2010). Depending shape which explains why cable vibrations decay with a linear

6
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

opt
Figure 7. Energy decay time for δ E = 98%: for all combinations of γ and c/c1 (a), for pure friction and pure linear viscous damping (b).

envelope far away from the anti-node position, as observed in 4. Numerical results
figure 6(a). The nonlinear vibrations result in both the time
variation and the position dependency of the point-to-point 4.1. Optimal damper parameters for minimum energy decay
damping ratios. Hence, although the logarithmic decrement time
is useful for the estimation of damping ratios of linear systems,
The simulation results indicate that an almost pure friction
the method cannot be used here. Therefore, the decay tests
damper with a γ ≈ [1.45–1.475] and negligibly small
are assessed by the time t E that is needed by the damper opt
viscosity of c/c1 ≈ [0–0.01] minimizes the decay time for
to reduce the total mechanical cable energy E tot down to a 98% cable energy reduction (figure 7). The optimal value γ
defined threshold δ E ((20), (21)). As can be seen from (20), the can only be given as a range for the following reasons:
total mechanical energy is divided by its value just before the
• The method used to estimate Ua in real-time shows an
decay starts in order make it non-dimensional. By doing this,
error (<1%) that has an impact on which value of γ turns
the damper performance assessment does not depend on the
out to be close to its optimum.
absolute value of cable amplitudes and therefore the damper
• All values of γ that evoke the minimum t E plus the
optimization does not depend on the chosen amplitude of the
maximum time step size of 1 × 10−3 s used for the
excitation force (13).
simulations must be denoted as close to the optimal value.
E tot (t = 140) − E tot (t) The solid line with circles in figure 7(b) displays the
δE = = [(0.1:0.1:0.9), 0.95, 0.98]
E tot (t = 140) energy decay time for differently tuned pure linear viscous
(20) dampers since the parameter γ tunes the damper viscous
t E (γ , c, δ E) = t (γ , c, δ E) − 140. (21) coefficient according to (7). This line shows the minimum at
γ = 1 which confirms the known optimal tuning of the damper
3.5. Optimization task viscous coefficient according to Krenk (2000) and thereby
validates the simulation tool. The contour lines for constant
The damper optimization parameters γ and c are sought to energy decay times in figures 8 and 9(a) indicate that, for
minimize the time t E needed to reduce the total mechanical decreasing energy decay range, the optimal damper parameters
cable energy E tot to the threshold δ E defined in (20). shift from predominantly friction damping to almost optimal
linear viscous damping with a very small friction part. This
(γ , c)opt for t E = min(t E ). (22) result is better visualized in figure 9(b) by the optimal
opt
combinations of γ and c/c1 as a function of δ E . It is seen
3.6. Variation ranges of damper optimization parameters from this figure that several optimal combinations of γ and
opt
c/c1 lead to the same energy decay time t E in the region
The increments, and lower and upper bounds of the damper
of 10%  δ E  30%. This contradictory outcome results
optimization parameters γ and c are chosen so that the
from the time-discontinuous energy dissipation of the damper
minimum of t E is clearly resolved and does not lie on the
that evokes the ripple seen in the time history of the total
bounds ((23), (24)). mechanical cable energy in figure 6(b). When evaluating the
range(γ ) = [(0.9:0.1:1.3), (1.35:0.025:1.5), decay of the total mechanical energy for t E , especially for
(1.55:0.05:1.8), (1.9:0.1:2)] (23) small δ E , the ripple may lead to equal values of t E for several
opt
opt combinations of γ and c/c1 . Neglecting this effect, the main
range(c) = c1 [(0:0.01:0.07), (0.1:0.1:0.6), tendency observed from figure 9(b) can be summarized as
(0.8:0.2:1)]. (24) follows.

7
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 8. Contour plots for δ E = 98%, 95%, 90% and 80% (a) and δ E = 70%, 60%, 50% and 40% (b).

opt
Figure 9. Contour plots for δ E = 30%, 20% and 10% (a); optimal combination of γ and c/c1 depending on the energy decay range δ E (b).

(a) If δ E  20%, a damper with characteristics close to those to the optimum. An explanation for this optimum is given in
of optimal linear viscous damping performs best. section 4.2.
(b) If δ E > 40%, a predominantly friction damper with
opt
c/c1 ≈ [0–0.01] leads to fastest cable energy removal. 4.2. Energy dissipation of friction dampers
(c) If 20% < δ E  40%, a combined viscous–friction
damper maximizes the energy decay rate. In the following, the impact of the optimization parameter γ
on the cycle energy of pure amplitude proportional friction
Another interesting result is obtained when the energy dampers is investigated. The power of the friction damper as
decay time histories due to optimal linear damping and pure a function of time Pfri (t) is the product of the damper velocity
friction damping are compared (figure 10). This figure plots and friction force
the cable energy averaged over one damper cycle E tot−cyc in
order to get rid of the ripple as visible in figure 6(b) and Pfri (t) = u̇ a (t) f (t). (25)
divided by the cable energy during steady state conditions
E ss before decay start in order to compare the energy decay Integration of the damper power over the time period Td of
time histories better. It turns out that optimal linear viscous the excited mode yields the cycle energy of the friction damper
damping outperforms pure friction damping at any γ during E fri as follows (figure 11(a))
the phase of reducing the first 56% of cable energy. After  t (k+1)
that, pure amplitude proportional friction damping performs E fri ((t(k+1) + t(k) )/2) = Pfri (t) dt. (26)
significantly better, and γ ≈ [1.45–1.475] seems to be close t (k)

8
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 10. Cable energy decay for pure friction damping depending on γ and optimal linear viscous damping as a benchmark.

Figure 11. Cable displacement at the damper position (a); ratio between damper cycle energies due to friction damping with constant γ and
optimal linear viscous damping of each mode (b).

As explained in section 3.3, the velocity independent viscous dampers to the cycle energy of the friction damper.
friction force induces the stop and go behaviour of the friction The optimal damper coefficient for mode n is given by (3).
damper and thereby excites the higher symmetrical modes (19) Considering that mode n generates n full damper cycles during
via energy spillover. Thus, band pass filtering of the cable the time period Td , the energy dissipated by the fictitious
displacement u(t, a) at the damper position yields the modal optimal linear viscous damper during Td becomes
displacements u n (t, a) of the symmetrical modes at the damper
E vis−n ((t(k+1) + tk )/2)
position (figure 11(a) shows the modal displacements up to    
n = 7) Un (t(k) , a) + Un (t(k+1) , a) 2
=n π 2π f r n cn .
0 opt
(29)
2
u(t, a) −−−−−−−−−→ u n (t, a),
band pass filtering
(n = 1, 3, 5, . . .). The fictitious energies due to optimal linear viscous
(27) damping are determined up to mode 15. Higher modes are not
At the same time instants t(k) , the modal displacement considered because they would generate less than 0.1% of the
amplitudes Un (t(k) , a) are selected. The sinusoidal nature of damper cycle energy due to mode 1. The fictitious, optimally
Un (t(k) , a) allows the derivation of the velocity amplitudes of tuned linear viscous dampers would act simultaneously on the
mode n as follows cable. Therefore, the sum E vis of their damper cycle energies is
the total energy removed from the cable during the time period
U̇n (t p,k , a) = 2π f rn0 Un (t p,k , a). (28)
Td
The idea is now to damp each modal component by a 
15
fictitious, optimally tuned linear viscous damper and then to E vis ((t(k+1) + t(k) )/2) = E vis−n ((t(k+1) + t(k) )/2). (30)
compare the cycle energies of these fictitious optimal linear n=1

9
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

Figure 12. Energy equivalence of an optimal friction damper (a); ratio between damper cycle energy due to friction damping with transient γ
and optimal linear viscous damping of each mode (b).

It is found that, during steady state conditions, the cycle Table 1. Optimal γ for min{t E (δ E = 98%)} and different damper
energy of the friction damper E fri is approximately equal to positions.
E vis if γ is chosen to be in the range of 1.45–1.475. This a/l (%) γ for min{t E (δ E = 98%)} r E (t = 135 s)
result is plotted in figure 11(b) using the ratio r E (31) of the
two damper cycle energies. 1 1.55–1.6 1.0864–1.0387
2 1.45–1.475 0.9952–1.0051
E fri ((t(k+1) + t(k) )/2) 3 1.45–1.5 0.9384–0.9645
r E ((t(k+1) + t(k) )/2) = . (31) 4 1.45–1.5 0.9982–1.0307
E vis ((t(k+1) + t(k) )/2) 5 1.487–1.525 0.9498–0.9792
6 1.475–1.525 0.9389–0.9812
It can be concluded that (figure 12(a)):
(a) friction damping is always multi-mode damping even
though only one mode is excited and This procedure was applied between 140 and 180 s. The
(b) optimal friction damping is equivalent to optimal linear resulting time histories of the transient γ and the energy ratio
viscous damping of each modal component. r E are displayed in figure 12(b). It turns out that the decay
time for δ E = 98% due to the transient γ is slightly larger
In order to test conclusion (b), free vibration decay tests at (t E = 19.23 s) than that resulting from the constant value of
different values of γ were also simulated for other reasonable 1.475 (t E = 18.96 s). It seems that transient energy spillover
damper positions, i.e. the damper was located at 1%, 3%, 4%, during the decay phase, which transfers an additional amount
5% and 6% cable length, since the damper position influences of energy from lower to higher frequencies for a short time
the tuning of the friction and linear viscous damper by (6) interval, increases slightly the energy dissipation rate in the
and (3), respectively. The decay tests were again evaluated friction damper.
by the time needed to reduce the cable energy to 2% of the
initial value. The approximate optimal γ values (ranges) that
minimize t E (δ E = 98%) and the according ranges of the 5. Conclusions
energy ratios r E are given in table 1. These numbers confirm
conclusion (b) except for the damper positions of 5% and 6% A combined viscous–friction damper located at 2% cable
where the mean values of the ranges of r E differ by more than length was optimized for maximum cable energy reduction
2% from 1. during a free vibration decay test. The damper optimization
parameters were the damper viscous coefficient c and the
proportional gain γ that adjusts the friction force in proportion
4.3. Transient tuning of a friction damper
to the cable amplitude at the damper position. This
The large deviations of r E from 1 mainly after 152 s proportional control compensates for the drawback of passive
(figure 11(b)) coincide with the time when transient energy friction dampers where the efficiency strongly depends on
spillover primarily takes place (figure 6(b)). Since γ amplitude. Since the feedback tunes the force to amplitude
determines the amount of energy spillover, a transient tuning ratio based on the last half period of vibration, the friction force
of γ should enable us to realize r E ≈ 1 for the entire decay still remains velocity independent. This nonlinear behaviour
time. The sought time history of γ was found by the variation generates energy spillover from the excited to higher modes,
of γ in time intervals of 2 s. If r E ([ti , ti + 2s])  1 ± 0.2 leads to almost rectangular cable displacement at the damper
was achieved, γ was varied within the next time interval etc. position and produces a rather triangular mode shape. It was

10
Smart Mater. Struct. 19 (2010) 045025 F Weber and C Boston

shown that, for this case, the logarithmic decrement method smart passive control strategy Struct. Control Health Monit.
used to estimate the damping ratio from free vibration decay 15 785–96
Christenson R E, Spencer B F Jr and Johnson E A 2006 Experimental
tests, and thereby to assess the damper performance, fails.
verification of smart cable damping J. Eng. Mech. 132 268–78
Instead, the damper efficiency was assessed by the time needed Ciǧeroǧlu E and Oz̈guv̈en H N 2006 Nonlinear vibration analysis of
for the damper to lower the total mechanical cable energy to a bladed disks with dry friction dampers J. Sound Vib.
defined threshold. 295 1028–43
The simulation results revealed that a damper with Fujino Y and Hoang N 2008 Design formulas for damping of a stay
cable with a damper J. Struct. Eng. 134 269–78
predominantly friction damping at γ ≈ [1.45–1.475] and very Guglielmino E, Edge K A and Ghigliazza R 2004 On the control of
opt
small viscous damping at c/c1 ≈ [0–0.01] produces the the friction force J. Mecc. 39 395–406
fastest cable energy reduction for the removal of 40% or more He W L, Agrawal A K and Yang J N 2003 Novel semiactive friction
of the total mechanical cable energy. If the goal is to lower controller for linear structures against earthquakes J. Struct.
Eng. 1291 941–50
the first 20% of the cable energy as fast as possible, almost
Huanga F-L, Wanga X-M, Chena Z-Q, Hea X-H and Nib Y-Q 2007
optimal linear viscous damping performs best, thus γ ≈ 1 A new approach to identification of structural damping ratios
opt
and c ≈ cn . For the fastest removal of 20%–40% of cable J. Sound Vib. 303 144–53
energy, the optimal damper parameters change rapidly between Inaudi J A 1997 Modulated homogeneous friction: a semi-active
the above values. damping strategy J. Earthq. Eng. Struct. Dyn. 26 361–76
Irvine H M 1981 Cable Structures (Cambridge, MA: MIT Press)
The simulation data were then analysed in order to explain Krenk S 2000 Vibrations of a taut cable with an external damper
the optimum around γ ≈ [1.45–1.475]. Since the friction J. Appl. Mech. 67 772–6
damper generates almost rectangular cable displacement at the Krenk S and Høgsberg J R 2005 Damping of cables by a transverse
damper position, the modal components of the excited mode force J. Eng. Mech. 131 340–8
and the modes due to energy spillover were derived by band Li H, Liu M, Li J, Guan X and Ou J 2007 Vibration control of stay
cables of the shandong binzhou yellow river highway bridge
pass filtering. Then, the cycle energies of fictitious linear using magnetorheological fluid dampers J. Bridge Eng.
viscous dampers, which were optimally tuned to each modal 12 401–9
component, were determined. It turned out that the sum of Main J A and Jones N P 2002 Free vibrations of taut cable with
these cycle energies is equal to the cycle energy of the friction attached damper. II: nonlinear damper J. Eng. Mech.
128 1072–81
damper at γ ≈ [1.45–1.475]. This has demonstrated that an Neelakantan V A and Washington G N 2008 Vibration control of
inherent feature of friction damping is multi-mode vibrations structural systems using MR dampers and a ‘modified’ sliding
and optimal friction damping is equivalent to the optimal linear mode control technique J. Intell. Mater. Syst. Struct. 19 211–24
viscous damping of each modal component. Sims N D, Holmes N J and Stanway R 2004 A unified modelling and
model updating procedure for electrorheological and
magnetorheological vibration dampers Smart Mater. Struct.
Acknowledgment 13 100–21
Sun L, Shi C, Zhou H and Zhou Y 2005 Vibration mitigation of long
The authors gratefully acknowledge the financial support of the stay cable using dampers and cross-ties Proc. 6th Int. Conf. on
Cable Dynamics (Charleston, SC, Sept. 2005) pp 443–50
Swiss Federal Laboratories for Materials Testing and Research Vader T S and McDaniel C C 2007 Influence of dampers on seismic
(EMPA), Duebendorf, Switzerland. response of cable-supported bridge towers J. Bridge Eng.
12 373–9
Wang X Y, Ni Y Q, Ko J M and Chen Z Q 2005 Optimal design of
References viscous dampers for multi-mode vibration control of bridge
cables J. Eng. Struct. 27 792–800
Bathe K-J 1982 Finite Element Procedures in Engineering Analysis Weber F, Distl H, Feltrin G and Motavalli M 2009 Cycle energy
(Englewood Cliffs, NJ: Prentice-Hall) control of MR dampers on cables Smart Mater. Struct.
Bhaskararao A V and Jangid R S 2006 Harmonic response of 18 015005
adjacent structures connected with a friction damper J. Sound Weber F, Feltrin G and Distl H 2008 Detailed analysis and modelling
Vib. 292 710–25 of MR dampers at zero current J. Struct. Eng. Mech. 30 787–90
Boston C, Weber F and Guzzella L 2009 Optimal semi-active Weber F, Feltrin G and Motavalli M 2005a Passive damping of
damping of cables: evolutionary algorithms and closed-form cables with MR dampers J. Mater. Struct. 38 568–77
solutions Smart Mater. Struct. 18 055006 Weber F, Feltrin G and Motavalli M 2005b Measured LQG
Bournand Y and Crigler J 2005 The VSL friction damper for controlled damping Mater. Struct. 14 1172–83
cable-stayed bridges. Some results from maintenance and Weber F, Høgsberg J and Krenk S 2010 Optimal tuning of amplitude
testing on long cables Proc. 6th Int. Conf. on Cable Dynamics proportional Coulomb friction damper for maximum cable
(Charleston, SC, Sept. 2005) pp 199–204 damping J. Struct. Eng. 136 123–34
Choi K-M, Jung H-J, Lee H-J and Cho S-W 2008 Seismic protection Ye S and Williams K A 2006 Torsional friction damper optimization
of base-isolated building with nonlinear isolation system using J. Sound Vib. 294 529–46

11

You might also like