CE8930 SHM GROUP REPORT 10
Group Number: 1
1
1.1
Feedback
Member 1
From the available resource, the quality of the error localization depends strongly on the number and
location of the sensors. I will look into the performance of ECRE method when there are only limited
measured DOFs data from the experiment. It could be helpful before we conduct an experiment test
on real structure. Additionally, I will simulate the effects of damage partially overlapping two DOFs
to determine if the ECRE method can still localize the damage.
In my previous simulations, the damage was only simulated by reducing Young’s modulus. I will try
to use some other ways to simulate the damage in finite element model. For example, we can adjust
the cross-sectional area or reduce the moment of inertia.
1.2
Member 2
Through the discussion of my individual project with my group, they felt a load cell would be
beneficial to verify the actual prestress force in the strand. Also, they thought hand calculations for
the stresses in the concrete would be beneficial to determine if any of the concrete’s cross section was
in tension due to the eccentricity of the prestressing force. If the top of the beam is in tension, it could
explain the lesser stiffness from the frequency response function in higher modes due to crack
opening. A final suggestion that I received from my group consisted of checking boundary
conditions. When the prestress is applied, the beam begins to camber, which may affect the boundary
conditions of the beam at the bearing points. On my next round of measurements, I will pay close
attention to the beam-ends above the neoprene pads to see if the beam is lifting off the pads.
1.3
Member 3
The group suggested that I evaluate my boundary conditions. Due to the size of my specimens, it
should be easy to modify them by placing a steel plate on one end and a steel tube on the other to
create a pin-roller connection. It would also be useful to perform hand calculations determining the
first natural frequency to verify modal software analysis calculations. My beams should also be
closely inspected to make sure that there is no damage, such as cracking, which may influence my
results. Also, the bar use to tension the threaded rods and prevent the rod from slipping (Figure 3)
should be closely monitored to ensure that it is not affecting a change in boundary conditions.
1
Group 1
2
Member 1 Report
The simply supported beam model was created in ANSYS. The whole beam was meshed into 50
elements. The damage is simulated by change the material property of the elements. Two types of
material property are defined. As shown in the following figures, the white bar is healthy part with
original young’s modulus (E = 2.068e11 Pa) while the red bar is with reduced young’s modulus
which indicates the damage.
For each damage scenario, the first 10 modes of the beam are generated. Only the modes with
vibration in the measured degree of freedom will be used for the further analysis. Then, the residual
energy values for elements in each mode are calculated. These residual energy will be normalized by
the total energy of the displacement field e.g uT*K*u. After that, I summed up normalized modelling
error indicator plots together and made it into one single plot for each damage scenario. The plots in
Figure 1 are the summed-up residual energy (modeling error) results for element from left end to the
right end. The damaged element number is also shown on the plot. From the simulation results, we
can conclude that the proposed ECRE method is able to identify the location of simulated damage in
finite element model.
For the further study, this method will be applied to a steel frame ANSYS model. The connection
damage will be simulated by reducing the rotational stiffness of spring elements. The expected results
are the proposed method is capable of localizing the damage in the element near the damaged
connection. Once the ECRE method can successfully identify the damage in the finite element model,
the experiment on the real steel frame will be conducted.
Simply supported beam with damage at elements 1-10
Simply supported beam with damage at elements 1-10
Figure 1 - Simulation results
2
Group 1
3
Member 2 Report
For my individual project, I have performed modal analysis using Pulse Hammer for various
prestressing levels. In total, 22 measurements were collected, with three accelerometers mounted
along the beam (Figure 2). One accelerometer was placed at mid span, and the other accelerometers
were placed at 18” on center from mid span in both directions. The hammer excitation is applied 2”
to the left of mid span. The measurements consisted of a zero prestress level, 0.25 kip prestress level,
and 1 kip increments up to 20 kip in the prestressing strand. The tests were performed on a relatively
mild day, from 12:00 to 4:00 pm with an estimated 10-degree temperature differential from
measurement 1 to measurement 22 (which could pose a few problems due to differing natural
frequencies with temperature differentials). For testing, I tested the prestressed beam up to 1.6 kHz,
with 6400 lines for ultimate precision. Seven averages were used for each measurement to attempt to
decrease noise. Ambient vibration in the system was a problem for these tests as the accelerometers
fluctuated around the threshold settlement values (blue to green increments on Pulse), which I have
never encountered before in previous tests, so further research will be needed to determine the source
of these ambient vibrations. The accuracy of the actual prestress force in the strand is around ± 0.1
kips, which I felt was adequate for this preliminary research. The frequency response functions and
coherence for all 22 measurements are presented below.
Figure 2 – Frequency response function for varying post-tensioning loads
For further study, features will be extracted from the 22 measurements, such as natural
frequencies/damping from Pulse Reflex, temporal moments, and possibly magnitudes. It should be
noted that if magnitudes are used a higher order of averages will need to be performed to reduce the
error due to noise. Once all features are extracted, a direct comparison will be obtained for these
features, and will provide me with information pertaining to the change of dynamic characteristics of
the system with additional levels of prestressing force.
3
Group 1
4
Member 3 Report
I completed hammer testing on my small
specimens with six distinct load levels. Loads
were applied using a threaded rod with a bearing
plate on the end of each specimen (Error!
Reference source not found.). The steel bar at
the back of the specimen is used to secure the rod
so that it does not turn as a tensile force is applied
to the threaded rod. This rod could be a potential
source of error but cannot be removed during
Figure 3 – Test setup
testing. To neutralize this issue, I have attempted
to maintain the rod is a similar position between each test to keep boundary conditions as consistent as
possible.
The results are encouraging and can be seen in Figure 4. The first mode is not very reflective of
change in the system, but the second and third modes show consistent shifts. These values are
summarized Figure 5. Additional testing needs to be performed using improved boundary conditions
as suggested in the group feedback. Further work will focus on evaluating data for relevant features
and creating metrics to compare applied post-tensioned loads.
0.6
Accelerance [(m/s2)/N]
0 lb
0.5
553 lb
0.4
1100 lb
1660 lb
0.3
2400 lb
3320 lb
0.2
0.1
0
0
200
400
600
800
1000
1200
1400
1600
1800
2000
Frequency (Hz)
Figure 4 – Frequency response function for varying post-tensioning loads
Frequency (Hz)
440
420
400
380
Mode 2
360
Mode 3
340
0
1000
2000
3000
4000
Load (lbs)
Figure 5 – First and second mode natural frequencies at varied post-tensioning loads
4
Group 1
5