You are on page 1of 6

Dynamic Testing of a Pre-Stressed Concrete Beam

Serge JACOBS1, Guido DE ROECK2


1
Ing, K.U.Leuven, Department of Civil Engineering, Structural Mechanics
Kasteelpark Arenberg 40, 3001 Heverlee
2
Prof Dr ir, K.U.Leuven, Department of Civil Engineering, Structural Mechanics
Kasteelpark Arenberg 40, 3001 Heverlee
E-mail: serge.jacobs@bwk.kuleuven.ac.be, guido.deroeck@bwk.kuleuven.ac.be

Abstract: A series of static load tests and dynamic tests where Only a few studies focused on one or more aspects of this
conducted on a pre-stressed (post-tensioned) concrete beam. project. Kato and Shimada [1] performed static and dynamic
Literature on modal changes due to damage of pre-stressed tests on a PC Bridge. They noticed a decrease in natural
concrete structures, is very scarce. One of the innovating aspects frequencies once cracks start to form. They state that as long
in this project is the introduction of Fiber Optic Sensors (FOS) to
as the prestressed steel wires don’t exceed their plastic limit,
measure the dynamic strains. Eight FOS are placed inside the
beam (in two cross sections) and an external FOS is used to cracks may close after the load has been removed and the
measure the strains at six different locations in the same two changes in eigenfrequencies will be remote. However, once
cross sections. Although a drop in eigenfrequencies is noticed the plastic limit is exceeded, weaknesses can be discovered by
with increasing load, significant shifts are only measured when vibration measurements. They also noticed that there are
the reinforcement starts yielding. Damping values tend to rise practically no changes in damping values. Maaskant et al [2]
with increasing damage. Results show however that the monitored the stress loss in the prestress tendons of a bridge
curvatures of modeshapes change with increasing load. Another using Bragg grating FOS. Abdalla and Kennedy [3] show in
result is the phase angle between the modal strains and the their paper that beams with openings have lower
vertical acceleration. Already in a very early stage (before any
eigenfrequencies than solid beams. When the opening is
cracks are visible) changes in the phase angle can be denoted.
Results from Direct Stiffness Calculation and FEM Updating are located at a node (zero displacement in a mode shape), it will
also shown. These results show a decrease in stiffness in the cause a sudden drop of the eigenfrequency associated with
damaged zones. The (asymmetrical) damage patterns coincide that mode shape. The mode shapes themselves are also
well with the one expected from visual checks. strongly influenced by the presence of openings, particularly
Keywords: Dynamic, Pre-Stressed Concrete, Modal Strains, near the openings themselves.
Mode shapes, Stiffness In the present research project static tests are performed to
induce damage in the beam in a controlled way. After every
static load, a series of dynamic tests are performed to
I. INTRODUCTION determine eigenfrequencies, mode shapes, damping values
and modal strains. This paper gives an overview of the
In a previous research project 5 reinforced concrete beams
dynamic test procedure and the results.
with dimensions 6 m x 0.2 m x 0.25 m where tested statically
and dynamically. It was concluded that eigenfrequencies are
II. DYNAMIC TEST PROCEDURE
good damage indicators. Changes of mode shapes are rather
small. On the contrary, modal curvatures or modal strains are For the dynamic test procedure, free-free conditions are
much more sensitive to damage but are difficult to measure obtained by lifting the beam on air cushions. The total
because of their very low amplitude. Deriving modal elevation is about 3 cm. To prevent the beam from moving in
curvatures from modal displacements is possible by the lateral and transversal direction and still allow bending in
differentiating twice, but this process is prone to measurement the vertical plane and torsional rotation, steel cables are used
and rounding-off errors. Therefore it would be beneficial to which are fixed to the beam.
measure directly modal strains. A drop weight excites the beam. Its weight is about 115 kg
The present research project, entitled “Enhanced and the falling height is about 1 m. The weight falls on a
performance of dynamic monitoring of civil constructions by damper that acts as a mechanical low-pass filter with a cut-off
integrating optical fiber sensors”, is focused on three main frequency of about 120 Hz. The weight is dropped at one end
objectives. First of all, the degradation of the stiffness of a of the beam, asymmetrical to the longitudinal axis in order to
pre-stressed concrete beam is studied. When the cracks excite torsional modes as well. Fig. 1 shows the entire
inducing load is removed from the beam, these cracks close dynamic test setup.
again. So the question arises if any changes in Vertical accelerations are measured in 75 different points,
eigenfrequencies and mode shapes can be observed. 37 points on each side of the beam and 1 reference point at
A second major achievement is the introduction of Fiber the left side of the beam in point 1000 (fig. 2). The
Optic Sensors (FOS) to measure strains dynamically. Bragg accelerations are measured with PCB piezoelectric
grating FOS where placed inside the beam and an external accelerometers of type 338B35 with a sensitivity of 100
Bragg grating FOS was developed to measure strains. mV/g.
Finally, the test results will be used for validation of
damage identification methods.
kept, which corresponds with a total time of 10 to 20 seconds.
The sampling frequency for all measurements is 256 Hz.
An analog low-pass filter is placed between force sensors;
internal and external strain sensors on one hand and the
Scadas III front-end. The cut-off frequency for the force is
128 Hz and for the strain sensors 80 Hz.

III. DYNAMIC TEST RESULTS


The eigenfrequencies, modeshapes and damping values are
calculated with SPICE (a system identification software,
based on the Stochastic Subspace Method, developed at the
K.U.Leuven) [4]. In the range from 0 to 128 Hz, 8 modes are
found, 4 torsion modes and 4 vertical bending modes. Fig. 4
Fig. 1. Dynamic test setup gives an overview of the modes in the initial state of the
beam.

Vertical bending modes Torsion modes

Fig. 2. Acceleration measurement points

Strains are measured in 14 locations. 8 Fiber Optic Sensors


(FOS) are placed inside the beam and an external movable
FOS is used to measure strains in 6 other locations. Strains are
measured in the mid-section of the beam and in the section at
5.2 m from the left (fig. 3).

Fig. 3. Position of FOS in the mid-section (left) and section


5.2 m (right)

The acquisition system used is a LMS Scadas III front-end


with 16 channels. This front-end is connected to a laptop with
the Cada-X software. Every measurement is divided in 8
setups. In every setup the impact force, the reference signal, Fig. 4. Overview of the modeshapes in the initial state of
an internal strain signal and an external strain signal is the beam
measured. The impact force is the added signal of 3 force
sensors that are placed beneath the damper. The remaining 12 As expected, the eigenfrequencies drop as damage
channels are used for the accelerations on the beam. increases. Fig. 5 shows the eigenfrequencies versus applied
For every setup signals are recorded continuously during 4 load for the first 4 bending modes. These curves only show a
minutes. This corresponds with an average of about 7 to 8 very significant decrease from a load of about 140 kN. This
impacts per setup. After pre-processing only 5 impacts are load is near the load where the passive reinforcement starts to
yield.
Bending Modes

100.000
mode 1

mode 3
Table 2. Measured damping values
90.000
mode 5
80.000
mode 7

70.000
Eigenfrequencies / Hz

60.000

50.000

40.000

30.000

20.000

10.000

0.000
0 20 40 60 80 100 120 140 160 180
An interesting change is the one noticed for the modeshapes
Q / kN
(all mode shapes are scaled to a unit value in reference node
1000 (fig. 2)). Fig. 7 shows mode 1 (1st symmetrical vertical
Fig. 5. Eigenfrequencies versus load for the bending modes bending mode) in its initial state and after the last load step.

Fig. 6 illustrates that as well. Here the relative changes in


eigenfrequencies (in percent) are plotted against the applied
load for the bending modes. Only after step 2 (67.5 kN:
cracks are already visible) a relative change is noticed. It has
been found from bridge tests that the 1% threshold
corresponds to the remaining uncertainty after elimination of
the influence of the environmental parameters such as
temperature [5].

Fig. 7. 1st bending mode Q = 0 kN (left) and Q = 154 kN


Relative changes eigenfrequencies bending modes

100.0 (right)

Fig. 8 shows the same for mode 3 (2nd vertical bending


10.0 mode, 1st anti-symmetrical vertical bending mode) and Figure
9 for mode 5 (3d vertical bending mode, 2nd symmetrical
df/f / %

Mode 1

Mode 3 vertical bending mode).


Mode 5
1.0 Mode 7
0 45 67.5 125 140 150 154

0.1
Q / kN

Fig. 6. Relative change in eigenfrequencies versus load for the


bending modes

Similar results are found for the torsion modes. An


overview of all the measured eigenfrequencies is given in
Table 1.
Fig. 8. 2nd bending mode Q = 0 kN (left) and Q = 154 kN
Table 1. Measured eigenfrequencies (right)

Table 2 shows that damping values slightly increase as


damage increases. Damping values are difficult to determine
though. Mode 6 (3d torsion mode) has higher damping values
than all other modes. This is probably due to the steel cables Fig. 9. 3d bending mode Q = 0 kN (left) and Q = 154 kN
that prevent the beam from moving laterally and transversally (right)
and apparently also influences slightly the torsional motion.
A clear deformation of the modeshapes can be denoted. The 20° or 160°. When the first cracks are visible the phase angle
zone of high curvatures is also the location where the first rises up to 60° or 120° and remains almost constant till
cracks were visible and where the beam finally broke down. yielding of the passive and active reinforcement. Then the
From fig. 7 it is clear that the modal curvature in the mid- phase angle drops/rises again. Similar results are found for the
section decreases at the ultimate load step. external strain sensor. The increase in phase angle is probably
One of the innovating aspects of this project is the dynamic due to debonding of the reinforcement in the concrete-matrix
measurement of strains by optical fibers. Fig. 10 shows the and imperfect fitting of the crack faces. On the other hand, the
modal strains for mode 1 in the mid-section for the internal modal damping factors don’t show a similar behavior (Table
FOS versus the applied load. A clear drop in the modal strains 2), they remain almost constant till the ultimate load steps.
can be denoted starting from a load of about 120 kN. Similar
results are found for the internal sensor in the section at 5.2 m Modal Strains Phase Angles

as well for mode 1 as for mode 3.


180

160

Internal, mid-section, mode 1


140
m1 50

0.03 120 m1 51

m1 52

Phase / deg
100
0.02 m1 53

80 m1 54

m1 55
0.01
60
m1 56

m1 57
0 40
Modal strain

0 20 40 60 80 100 120 140 160 180


20
-0.01
0
0 20 40 60 80 100 120 140 160 180
-0.02
Q / kN

m1 50
-0.03
m1 51

-0.04
m1 52 Fig. 12. Modal strain phase angle versus load for the internal
m1 53
Q / kN FOS

Fig. 10. Modal strains of mode 1 in the mid-section for the IV. DIRECT STIFFNESS CALCULATION (DSC)
internal FOS versus load The DSC method [6] uses modal curvatures to calculate
decrease in stiffness.
Fig. 11 shows the comparison between the measured modal
strains in the mid-section before loading and after the last load K * ϕm = ωm² * M * ϕm (1)
step. The measured modal strains before loading correspond
well with the ones calculated by the ANSYS FE-program.
Where K is the (unknown) stiffness matrix, ϕm is the vector
After the last load step the modal strain curve is much flatter.
of measured modal displacements, ωm is the measured
Similar results are found for the internal sensor in the section
eigenfrequency and M is the (known assumed) mass matrix.
at 5.2 m and for the external sensor.
ωm² * M * ϕm represents the inertial load. With this known
(measured) inertial load, modal bending moments (M) and
modal torsional moments (T) can be calculated. This leads to
mid-section

0.0300 the following formulas:


Ansys

For bending modes: EI = M / (d²ϕmb/dx²) = M / κb (2)


0.0200 Q0

Q154

0.0100 For torsion modes: GJ = T / (dϕmt/dx) (3)


Modal strain

Where ϕmb are the modal bending displacements, ϕmt the


0.0000
0 10 20 30 40 50 60 70 80

-0.0100 modal torsional angles of twist and κb the modal curvature.


The DSC has only been applied for bending modes. Fig. 13a
shows the modal curvature of mode 1 versus the length of the
-0.0200

-0.0300 beam for all load steps. There is a clear increase of the modal
y / cm
curvature for load steps 5 and 6 and in a lesser extend in load
step 4 around 10 m from the left side of the beam. Fig. 13b
Fig. 11 Comparison of modal strains versus distance to shows the distribution of the stiffness EI along the beam for
bottom of the beam at Q = 0 and Q = 154 all load steps. A significant decrease in stiffness, which
coincides with an increase of damage, is noticed around 10m
Interesting results are shown in fig. 12. The phase angle from the left side of the beam.
between the modal strains of mode 1 and the vertical Fig. 13c shows the modal curvature of mode 3 versus the
acceleration in the reference point is plotted against the length of the beam for all load steps. Again, a clear increase in
applied load. Depending on the location of the measurement modal curvature is noticed for load steps 5 and 6. Fig. 13d
point (above or under the center of gravity) the initial phase shows the distribution of the bending stiffness EI along the
angle is 0° or 180°. At Q = 45 kN the phase angle is already beam for all load steps corresponding to mode 3. Although
the middle zone cannot be identified (due to zero by zero The damage pattern is very similar to the one calculated by
division), similar conclusions about bending stiffness the DSC method. Again we find an asymmetrical damage
reduction can be drawn as in the case of mode 1. pattern with the highest damage at about 10 m from the left
side of the beam.

a. Modal curvatures κb b. Bending stiffness EI


versus length of the beam versus length of the beam
for all load steps for mode for all load steps for mode a. Distribution of the b. Distribution of the
1. 1. bending stiffness along the bending stiffness along the
beam for the initial state. beam after the first load step.

c. Modal curvatures κb d. Bending stiffness EI


versus length of the beam versus length of the beam c. Distribution of the d. Distribution of the
for all load steps for mode for all load steps for mode bending stiffness along the bending stiffness along the
3. 3. beam after the second load beam after the third load step.
Fig. 13. Modal curvatures κb and bending stiffnesses EI step.
versus length of the beam for all load steps for modes 1 and 3

V. FINITE ELEMENT MODEL UPDATING


A finite Element Model (FEM) of the test beam has been
updated during successive load steps in order to calculate the
decrease in bending stiffness.
The method uses “damage functions” [7] in order to reduce
the number of unknowns in the updating process. The beam is
assumed to be homogeneous and isotropic. A Poisson ratio of
0.2 is assumed. The passive reinforcements and the e. Distribution of the f. Distribution of the
prestressing cable were not modeled as such, but an bending stiffness along the bending stiffness along the
equivalent Young modulus was calculated to take the beam after the fourth load beam after the fifth load step.
reinforcement and pre-stressing cable into account. The step.
Young modulus used is 3.7855 Pa. Fig. 14. Distribution of the bending stiffness along the beam
The eigenfrequencies and modeshapes measured must for all load steps
coincide with the modeled ones. The bending stiffnesses are
adapted for optimum correspondence, in order to detect local
damage, as local damage will cause with a local decrease in VI. CONCLUSIONS
stiffness. From the dynamic tests following conclusion are made:
The initial model is brought in agreement with the · The eigenfrequencies drop with increasing damage,
measurements of the beam in its undamaged state (fig. 14a). It but a significant drop is only noticed once the passive
can be noticed that there are some differences between the reinforcement starts to yield.
initial FE model and the one after updating to the measured · Changes in the mode shapes are also noticed. The
state. This is mainly due to some initial shrinkage cracks. modal curvature lowers and a shift towards the zone of
Figs. 14b, 14c and 14d show that the relative changes in maximal damage occurs. But again, significant differences are
bending stiffness along the beam are small. Clear changes in only noticed once the passive reinforcement starts to yield.
bending stiffness are only observable after yielding of the · Damping values stay more or less constant during
passive (and active) reinforcement (figs. 14e and 14f). the entire test. A very slight increase is noticed after the
passive reinforcement start to yield, but this increase is not
significant enough.
· Changes in the phase angle between the horizontal
acceleration in the reference point and the horizontal modal
strains are also noticed. Significant changes are already
measured at a very early stage. Although further study on this
must be performed, this might indicate that this phase angle
can be used as a good damage indicator.
From the DSC and the FEM updating following
conclusions are made:
· Significant changes are only noticed once the passive
reinforcement starts yielding.
· Both methods show a decrease in bending stiffness
around the same area where the biggest cracks where noticed
during the testing. The damage pattern is asymmetrical.
· Both methods should be used in combination with
other methods in order to identify damage in an early state.

VII. ACKNOWLEDGEMENT
The current project G.0266.01 is supported by the FWO
(Fonds Wetenschappelijk Onderzoek) Vlaanderen.

REFERENCES
[1] Kato, M. and Shimada, S., (1986), Vibration of PC Bridge during
Failure Process, Journal of Structural Engineering, Vol. 112, No.7.
[2] Maaskant, R., et al. (1997), Fiber-optic Bragg Grating Sensors for
Bridge Monitoring, Cement and Concrete Composites, Vol.19, pp.21-
23.
[3] Abdalla, H. and Kennedy, J., (1995), Dynamic Analysis of
Prestressed Concrete Beams with Openings, Journal of Structural
Engineering, Vol.121, No.7.
[4] Peeters, B., (2000), doctoral thesis, K.U.Leuven, System
identification and damage detection in civil engineering.
[5] Maeck, J. and De Roeck, G., (2003), Damage assessment using
vibration analysis on the Z24-bridge, Mechanical Systems and Signal
Processing, V.71, No.1, pp.133-142, January 2003.
[6] Maeck, J., (2002), doctoral thesis, K.U.Leuven, Damage Assessment
of Civil Engineering Structures by Vibration Monitoring.
[7] Theugels, A., Maeck, J. and De Roeck, G., (2002), Damage
assessment by FE model updating using damage functions, Computers
and Structures, 80(25), pp.1869-1879, November 2002.

You might also like