presented by margaret g. brier, ozzie gooen, andrew ho, and sara sholes may 6, 2010

E80 Field Experience Engineering DepartmentHarvey Mudd College

Presented by

Margaret G. Brier, Ozzie Gooen, Andrew Ho, and Sara Sholes

May 6, 2010

E80 Field ExperienceNDE and System Identification of

a Concrete Bridge


Table of Contents Introduction

Background Statement of Work

Set-up Bridge Description & Configuration Measurement Layout

Instrumentation Accelerometer Matlab GUI NI DAQ Hammer and Tips Hammer Tip Selection

Testing Procedure Parameters Impulse Triggered Number of hits/trials

Data Processing Sample Data Description of Analysis Procedure PreFreq80 Data Processing Freq80

Interpretation of Results Response Frequencies and Shapes Damping Technical Highlight

Summary Appendixes:

Appendix A: FRF Plots at all Locations Appendix B: FRF Effects of Detrending and Windowing Data Appendix C: Heavy End Detrending


Background

Studies in the 1990s indicated the need to retrofit the nation’s bridges.

Non-destructive testing was implemented to systematically analyze these structures.

We have studied the Mountain Avenue Bridge, over the California 210 Highway.

The Mountain Avenue Bridge was designed in 1998 by the California Department of Transportation.


Statement of Work

From this analysis, we plan on identifying the: Fundamental Resonance Fundamental Response Shape Damping estimate

The fundamental resonance frequency is the frequency at which the bridge will oscillate at its maximum magnitude. At the first resonant frequency, the bridge’s response shape will be in the form of one period of a sine wave. If possible, we were to investigate the response shapes at higher modes. After determining the fundamental resonance, a damping estimate for the fundamental response can be found.


Bridge Description and Configuration


Measurement Layout

0 1 2 3 4 5 6 7 8 9

To take data along the length of the bridge, we placed two accelerometers as seen below and took ten sets of impact data at Locations 0 to 9 as shown.

Accel 1 Accel 2


Measurement Layout (con’t)

When choosing how many locations to impact, it was necessary to consider both quality of data and time constraints. We took as many data points as possible in the available time.

No data was taken when cars, pedestrians, or bicycles were moving across the bridge. This restricted the quantity of data.

In addition to the ten evenly spaced locations, data was also taken at the center of the bridge, on lamposts, around a joint, and on the guardrail.


Measurement Layout (con’t)

55.1’ 30.6’

275.6’

N

S

EW

1.0’

4.6’

0

1

2

3

4

5

6

7

8

9

Noacc

Soacc

•Testing performed solely on East walkway

•Accelerometers 1 ft from railing

•Impact testing also 1 ft from railing, along line of accelerometers


Instrumentation

The following were used to take data:

Accelerometer (Dytran Model 3191A1) Signal Conditioners/Filters Matlab based GUI with National Instruments DAQ

Center Calibrated Impact Hammer (Dytran Model 5802A) Hammer Tips (Lixie 200)


Instrumentation-Accelerometer

[2] http://www.dytran.com/products/3191A.pdf


Instrumentation-Signal Conditioner/Filter

[3] http://www.dytran.com/products/4105.pdf


Instrumentation - Matlab GUI

Force Impulse Channel

Accel 1 Channel

Accel 2 Channel

3 Channels Combined

Bonus Channel

Parameter Settings


Instrumentation - NI DAQ

[4] http://www.ni.com/pdf/manuals/321183a.pdf


Instrumentation - Hammer and Tips

[1] http://www.dytran.com/products/5802A.pdf

[2] http://www4.hmc.edu/engineering/eng80/lects/E80FE_FSSID_2010.pdf


Hammer Tip Selection

The Ideal tip should provide: pure impulse force minimal rise time zero force before and after impulse



Red tip

Black tip

Orange tip

Green tip


Tip Testing Conclusions

• Both frequency domain and time domain data show that the green tip was the best choice.• Our hammer tip analysis would have been more complete if the sampling resolution were higher.


Table of Contents Background Statement of Work Bridge Description & Configuration Measurement Layout Instrumentation

Accelerometer Matlab GUI NI DAQ Hammer and Tips Hammer Tip Selection



Interpretation of Results Damping Technical Highlight Summary Appendix A: FRF Plots at all Locations Appendix B: FRF Effects of Detrending and Windowing Data Appendix C: Heavy End Detrending


Testing Procedures

Figure. A Block Diagram of the Impact Testing Procedure


Parameters

• 4000 samples per second• 8 seconds total

• .2 seconds pre-trigger• 7.8 seconds post-trigger

• Trigger level = 1V above noise level• 25 Hz Filter


Impulse Trigger Method

Location 1, Trial 0, 4/20/10

ParametersImpulse Triggered

Number of hits/trialsRepeatabilitySaturation


Number of hits/trials3 hits processing


Sample Data

• Hammer Gain x1

• Accelerometer Gain x10• 25 Hz Cutoff

Location 1, Trial 0


Description of Analysis Procedure

Before Processing After Processing

•Windowing•Detrending•Removing Noise


PreFreq80 Data Processing

Force Impulse

Before Processing After Processing



• Detrending removes the best fit line

Force Impulse


• Windowing removes remaining noise

PreFreq80 Data Processing Force Impulse



• Close up on noise windowing

Force Impulse


PreFreq80 Data ProcessingAcceleration Processing

Before Processing

AfterProcessing


PreFreq80 Data ProcessingAcceleration Processing

Subtract mean from pre-trigger

Matlab detrend function with breakpoints in transient region

Detrend post-transient post trigger

Shorten pre-trigger(.2 seconds to .01seconds)

[200 1056 1250 1320:3000:16384]


Freq80

• Freq80 yields , an estimate of

• Assumes no noise

• Used block averaging

• Also Assumes periodicity• Needed to apply an exponential window

• Works best with minimal pre-trigger data.


Freq80

• Freq80 applied an exponential window using τ = .899 for a desired 1% of original signal by T = 4.096 seconds.


Data InterpretationSample Gain, Phase, and Coherence Data, after Freq80 (Location 1).


Bridge CharacteristicsClose-up of Gain with Coherence (Location 1).


Lampposts

When analyzing lamppost data, we see a peak at 3.4 Hz. 3.4 Hz peaks can be seen throughout the full bridge data.


Bridge Joint

These are examples of data from impacting around the joint between the bridge and the ground on the other side. There are no discernable resonances from the data around the joint.


GuardrailClear resonant frequencies can not be identified from the data taken from the guardrail.


Resonance Shapes

Resonance shape at 5.1 Hz.

We can see from this video that at 5.1 Hz the resonance shape resembles what we expect for the fundamental resonance.


Bad Resonance Shapes

• 8.8 Hz• No clear resonance shape


Bad Resonance Shapes

• 12.2 Hz• No clear resonance shape


Damping Estimate

Δ F

•Bandwidth appears at 3dB below resonant peak.

•Use quality factor Q relation.

•Q = fr / Δ F

•Q > ½ = underdamped

•ζ = 1/2Q = Δ F/ 2 fr

•Combined averages from all visible peaks.

•Damping Estimate ζ = .05


Technical Highlight CEMACZ: We compared how a theoretically predicted frequency response function

compared with the experimental FRF for a step and sinusoidal input. Dynamic Beam Modeling: When theoretically modeling the system, we considered the

response as a function of location (through static deflection) and the response as a function of time (through energy considerations)) separately, and then combined to determine the overall frequency response function. The theoretical input was an impulse.

Dynamic Beam Testing: We experimentally determined the frequency response function of the two distinct elements of the system (the beam and the TVA) based on the input and output signals, and from these responses designed the system to obtain desired overall frequency response.

Bucket Lab: We had a rough theoretical model for system, but did not know direct effect of various parameters. We determined some the parameters based on the log decrement displacement for a step input.

Wind Tunnel: In the wind tunnel, we did not treat the system as a 2nd order system or characterize a frequency response function. Instead, we used the Reynold’s number relation to design a system with the desired output.

Static Motor: We characterized the system based on the input and output data. The input data to this system was random vibration (which in the frequency domain is like a Gaussian distribution)


Technical Highlight

Generally, want to be able to understand the response of a system (might want to control damping, resonant frequency, bandwidth, etc…)

In E80, we explored various methods to characterize the response of a system


Summary

The experimentally determined fundamental resonant frequency of the bridge is 5.1 Hz.

The damping ratio is ζ=.05.


Thanks

The E80 Team:

Professors Zee Duron, Nancy Lape, Liz Orwin, and Qimin Yang

The Section 2 E80 Proctors:

Ariel Berman, Elizabeth Ellis, and Allie Russell Willie Drake and Sam Abdelmuati for preparing

and testing the instrumentation


(more) Questions?


Appendices

Appendix A: FRF Plots at all Locations Appendix B: FRF Effects of Detrending

and Windowing Data


Appendix A – FRF Plots at all Locations

Location 0



Location 1



Location 2



Location 3



Location 4



Location 5



Location 6



Location 7



Location 8



Location 9



Location Pier Support


Appendix B – FRF effects of Detrending and Windowing Data

Location 1, Trial 0, Test (a), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)



Location 1, Trial 0, Test (a), noacc Modifications:•Data sets have been reduced to block size 16384. (unprocessed)



Location 1, Trial 0, Test (b), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended, but not windowed



Location 1, Trial 0, Test (b), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended, but not windowed

Fully Processed Current Modifications



Location 1, Trial 0, Test (c), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended, and windowed to remove noise



Location 1, Trial 0, Test (c), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended, and windowed to remove noise.




Location 1, Trial 0, Test (d), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has noise removed, but not detrended



Recap: Modifications to Force Input

No detrend/window Detrend only




Detrend + WindowWindowing only




Comments:•Detrending only: Increases gain of FRF, introduces low frequency content•Windowing only: Little to no change•Detrend + Window: Reflects changes from both above

Next:•Reducing the pretrigger to .01 seconds rather than .2 seconds

• Better fits exponential windowing in Freq80



Location 1, Trial 0, Test (e), noacc

Modifications:•From Test(d), keep force detrended and windowed•Reduce pretrigger to .01 seconds



Location 1, Trial 0, Test (e), noacc

Test(d) .2 seconds pretrigger Test(e) .01 seconds pretrigger

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended/ windowed to remove noise•Pre-trigger time has been reduced to .01 seconds.



Location 1, Trial 0, noacc

Test(f) .1 seconds pretrigger Test(g) .15 seconds pretrigger





Recap : Reducing Pre-trigger Time

Comments:•After detrending/windowing the force input, reducing the pre-trigger time removes low frequency content, and increases visibility of several peaks in the frequency spectrum.



Location 1, Trial 0, Test (h), noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended and windowed to remove noise•Pre-trigger has been reduced to .1s•Pre-trigger acceleration response detrended



Location 1, Trial 0, noacc

Exponential region

Transient Region


Appendix B – FRF Effects of Detrending and Windowing Data

Location 1, Trial 0, Test(i) noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended and windowed to remove noise•Pre-trigger has been reduced to .1s•Pre-trigger acceleration response detrended•Testing various breakpoints to detrend transient and exponential portions of acceleration response

[200 1056 1250 1320:3000:16384]

3 points in transient, every 3000 in exponential



Location 1, Trial 0, Test(i) noacc

Modifications:•Data sets have been reduced to block size 16384. (unprocessed)•Force has been detrended and windowed to remove noise•Pre-trigger has been reduced to .1s•Pre-trigger acceleration response detrended•Testing various breakpoints to detrend transient and exponential portions of acceleration response

[0 813 865 941 1026 1122 1227 1280 1320:1000:16384]

Heavy end of detrending


Actually causes peaks to show up in south accel, at the cost of coherence[0 813 865 941 1026 1122 1227 1280 1320:1000:16384]

Heavy end of detrending


Compare with actual data used in total processing


Location 0

Appendix C- Heavy End Detrending