Structures/Motion Lab 20-263-571, Sections 001, 002, 003

STRUCTURAL BEAM TEST LAB OBJECTIVE:

The objective of this test is to conduct a vibration test on a simple structure (free-free beam) in order to determine the damped natural frequencies r and mode shapes r of the beam. Damping r ( r = r + j r ) will be present in the beam (material damping) but will be assumed to be small and constant for each mode. For the purposes of preparing for the lab, assume that the the beam will be made of aluminum with dimensions 3.0 inches wide by 36.0 inches long by 0.75 inches thick.

BACKGROUND/THEORY:
The simplest way to nd the damped natural frequencies and mode shapes of a simple structure is to recognize that the damped natural frequencies are approximately the frequencies where the frequency response functions reach a relative maxima. This is demonstrated by the following Figures.
Frequency Response Functions Magnitude 80

70

60

50

Magnitude

40

30

20

10

100

200

300 Frequency (Hz.)

400

500

600

Figure 1. Damped Natural Frequencies at Maxima of FRF Magnitude

+ -1Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Frequency Response Functions Magnitude Damped Natural Frequency

60

50

40

Magnitude

30

20

10

0 220 230 240 250 260 270 Frequency (Hz.) 280 290 300

Figure 2. Damped Natural Frequencies at Maxima of FRF Magnitude As long as the modal frequencies are not too close together, this simple approach will work well. The mode shapes can be found at these modal frequencies by understanding the Expansion Theorem. Expansion Theorem - Frequency Domain: { X( i ) } =

r=1

r { r }

The Expansion Theorem states that the response of the system is a linear combination of the mode shapes. At a specic damped natural frequency, the r coefcient for that mode shape will dominate. Thus, since r is a constant for each frequency, the response at each measurement point will be proportional to the mode shape. Since the frequency response functions are normalized response functions, the values of the frequency response functions, at the damped natural frequencies, will be approximately equal to the mode shapes. For a frequency response function, evaluated at the damped natural frequency r , the relationship is: { H( r ) }
pr

j 2 r Mr r

{ r }

This idea is demonstrated by the following Figures.

-2-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Frequency Response Functions Real Part 40

30

20

10

Real Part

10

20

30

40

50

100

200

300 Frequency (Hz.)

400

500

600

Figure 3. Damped Natural Frequencies - FRF Real Part

Frequency Response Functions Real Part 30

20

10

Real Part

10

20

Damped Natural Frequency

30

40 235 240 245 250 255 260 Frequency (Hz.) 265 270 275 280

Figure 4. Damped Natural Frequencies - FRF Real Part

-3-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Frequency Response Functions Imaginary Part 80

60

40

20

Imaginary Part

20

40

60

80

100

200

300 Frequency (Hz.)

400

500

600

Figure 5. Damped Natural Frequencies - FRF Imaginary Part

Frequency Response Functions Imaginary Part

60

40

20

Imaginary Part

20

40 Damped Natural Frequency 60

230

240

250

260 Frequency (Hz.)

270

280

Figure 6. Damped Natural Frequencies - FRF Imaginary Part

-4-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

This simple single degree of freedom (SDOF) approach assumes that the modal frequencies are sufciently separated in frequency so that the modal vector can be approximated by the response vector with no contamination (error) from modes that are lower or higher in frequency. The following Figure illustrates how single degree of freedom contrabutions combine to create the realistic multiple degree of freedom (MDOF) frequency response function. Note that when the natural frequencies are close together, the dashed line (SDOF) is not identical with the solid line (MDOF).

10 0 -10

Log Magnitude, (dB)

-20 -30 -40 -50 -60 -70

4 Frequency, (Hertz)

0 -20 -40 -60

Phase, (Degrees)

-80 -100 -120 -140 -160 -180 0 1 2 3 4 Frequency, (Hertz) 5 6 7 8

Figure 7. SDOF Contributions to Frequency Response Function Fur ther theoretical details are included in Appendix B.

-5-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

PROCEDURE:
The testing procedure will utilize a single reference, impact test approach to estimate the frequency response functions. This approach uses a single xed accelerometer as the response (output) sensor and an instrumented hammer as the excitation (input). The frequency response functions will be calculated by the Zonic Medalion Signal Analyzer.

The load cell that is permanently mounted in the impact hammer is to be connected to Channel 1 of the Zonic Medallion Signal Analyzer. The measurement sequence will be triggered on Channel 1 with a small pre-trigger delay (1-3% of the total block). A force window (rectangular) should be applied to the force signal to minimize noise problems. The impact hammer should have a nylon or plastic tip appropriate for the frequency range of the modes of the test structure. The accelerometer is connected to Channel 2. A response window (exponential) should be applied to the accelerometer signal to minimize the truncation error known as leakage. The accelerometer should be mounted with wax on the beam. Typical views of the Analyzer Setup and a typical views of the time domain histories (force and response) and frequency domain measurements (frequency response function and coherence) are included at the end of this lab documentation as Appendix B. Be sure to choose the frequency range (Bandwidth) large enough to nd the highest frequency mode of interest. Be sure to choose a Frame Size (number of time domain data points) of at least 1024. Free-free boundary conditions are to be approximated by resting the beam on a foam pad. Damping will be small enough to ignore. The stiffness of the foam pad may alter the free-free modal frequencies estimates by a ver y small amount. The location of the reference accelerometer and the location of the impact points is at the discretion of each lab group. Use 3-5 averages for each measured frequency response function. Be sure to turn on Manual Rejection for each average so that you can reject any average that has a double impact on the force signal or has an overloaded signal on either the force or response signal. Utilize the coherence function (measure of linearity between the measured input and output) to determine the quality of the measured frequency response functions. Save the measured frequency response functions to disc and move the data to MATLAB for plotting and/or processing. All damped natural frequencies and mode shapes should be identied before leaving the lab.

-6-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

RESULTS:

PRIOR TO COMING TO LAB, the damped natural frequencies for the rst six (6) deformation modes should be estimated analytically. Several references are available that provide the analytical solution for a simple free-free beam. For example, see Reference 1 or 2. Note that the rst six deformation modes may include bending or torsion modes involving any orientation of the rectangular beam. Experimentally determine the damped natural frequencies and associated mode shapes for the rst six deformation modes of the beam. Compare these values to your predictions. Be sure to note that your experimental data has a nite frequency resolution. Plots of the mode shapes should be included in the report along with a verbal description of each mode. The measured frequency response functions should be saved and moved to MATLAB to simplify the plotting of the mode shapes. Be sure to measure and weigh the beam to verify your analytical estimation.

DISCUSSION
The discussion should include the following issues:

Summarize the procedure used to identify the damped natural frequencies and mode shapes. Comment on any difculties encountered in identifying the mode shapes. Compare/contrast damped natural frequencies with theoretical values estimated from a reference. Discuss the reasons for any differences (what are the actual boundar y conditions?). Which answer is correct?

REFERENCES
1. 2. 3. Blevins, Rober t D., Formulas for Natural Frequency and Mode Shapes , Van Nostrand Reinhold Company, 1979, 492 pp. (Call Number TA654.B56) Volterra, Enrico, Zachmanoglou, E.C., Dynamics of Vibration , Charles E. Merril Books, Inc. 1965, 622 pp. (pp. 293-366) HP-35660 Dynamic Signal Analyzer, Getting Started Guide , The HewlettPackard Company, Manual Number 35660-90005, 1988.

-7-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

APPENDIX A: SDOF EXPERIMENTAL MODAL ANALYSIS CONCEPT

Figure 8. Modal Analysis Concept - Uniform Beam Even though a contiuous beam has an innite number of modes, the evaluation of these modes (estimation of frequency, damping, modal vector and modal scaling) can often be accomplished with essentially single degree-of-freedom (SDOF) concepts. The above diagram explains how this will be done. The primar y assumption is that each mode of vibration is well separated in frequency from the other modes. This is often the case for lightly damped structures. Different modes of vibration of the beam can be visulaized in the above gure by noting the solid black line connecting the peaks of the imaginary par ts of each frequency response. Normally, these modes are plotted in a wireframe model showing the extrema of the modal vector so that the motion can be easily understood. Figures 2 and 3 show the rst two bending modes of a uniform beam that is pinned at each end.

-8-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Figure 9. First Bending Mode at First Damped Natural Frequency ( 1 )

Figure 10. Second Bending Mode at Second Damped Natural Frequency ( 2 ) In order to understand why this information can be determined from the imaginary par t of the frequency response functions, SDOF theory must be reviewed and extended slightly, primarily from a notational point of view. The general mathematical representation of a single degree of freedom system is expressed using Newtons second law in Equation 1: M x(t) + C x(t) + K x(t) = f (t)
(1)

For the general case with a forcing function that can be represented as a summation of sin and cosine terms, the forcing function can be represented as: f (t) =

F( ) e j t + F * ( ) e j t =0
-9-

(2)

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Assuming that the system is underdamped and that enough time has passed that any transient response of the system due to initial condition or startup of the excitation has decayed to zero, the response of the system can be represented as: x(t) =

X( ) e j t + X * ( ) e j t =0

(3)

Note that, while x(t) and f (t) are real valued functions, X( ) and F( ) are complex valued. Working with any arbitrary frequency term in Equations 2 and 3, Equation 2, Equation 3 and the derivatives of Equation 3 can substituted into Equation 1 yielding the following frequency response function (FRF) relationship for a SDOF system: H( ) = X( ) 1 = F( ) M 2 + C j + K
(4)

Note the characteristic of the above frequency response function when it is evaluated (measured) at the undamped natural frequency. At the undamped natural frequency, the mass and stiffness terms cancel each other and the FRF is purely imaginary valued. The rst extension that is necessary provides a description for the case where x(t) and f (t) are not located at the same point. On a single degree-of-freedom system, this would provide redundant information (no new information) but it becomes important as the extension to multple degrees-of-freedom occurs. For example, assume that the par ticular point (and direction) on the mass where the force is applied is referred to as DOF p and the particular point (and direction) on the mass where the response is measured is referred to as DOF q. Equation 4 now can be written as follows to note this information. H qp ( ) = X q ( ) 1 = F p ( ) M 2 + C j + K
(5)

The system is still a SDOF system so H qp = H pp = H qq = H qs = . . . but the input and output location can now be described. This clearly demonstrates that the number of modes (one in this case) is unrelated to the number of input and output sensors that are used to measure the system. The second extension that is necessary provides a way to indicate that the modal characteristics (modal coefcients) of both the input and output are represented in the frequency response function model. The modal frequency is already represented by noting that the denominator is related to the characteristic equation. A form of modal scaling is already represented by noting the the mass term in the denominator scales the equation. Modal coefcient information, which is relative not absolute information, can be added by changing the numerator to reect this. H qp ( ) =
q p X q ( ) = 2 + C j + K F p ( ) M
(6)

-10-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Note that, since the system is still a SDOF system, the relative motion at each DOF would be normalized to 1 such that p = q = s = 1 which shows that Equation 5 and 6 still represent the same information. Note as before, if the FRF is evaluated (measured) at the undamped natural frequency), the FRF is once again imaginary valued and is a function of the modal coefcients and damping. Assuming the damping is unknown but constant means that the product of the modal coefcients is proportional to the imaginary par t of the FRF. Finally, the third extension that is necessary provides for the change from SDOF to MDOF. Note that for a linear system, linear superposition can be used in the frequency domain to add the information associated with each mode together to represent the frequency response function of a MDOF system. To desribe this, every term in Equation 6 will need a subcript (r) to indicate which mode the information is associated with. The nal form of the frequency response function is: H qp ( ) =
qr pr X q ( ) = 2 + C j + K F p ( ) r=1 M r r r
(7)

Equation 7 is one common representation of the FRF of a MDOF system. Note that the M r , C r and K r terms in the denominator are the modal or generalized mass, damping and stifness parameters, not the physical mass, damping and stiffness parameters. The modal or generalized parameters can be found analytically from the physical mass, damping and stiffness parameters or experimentally using more complicated parameter estimation algorithms. Note that, as long as the modes are well separated in frequency, the information in the neighborhood of the undamped natural frequency for a given mode can be found from: H qp ( ) =
qr pr X q ( ) 2 + C j + K F p ( ) M r r r
(8)

This characteristic is observable in Figure 4. Note that the SDOF contribution to the FRF (dotted line) is nearly the same as the MDOF contribution to the FRF (solid line) in the neighborhood of the undamped natural frequency, as long as the natural frequencies are well separated. Note that, if the output DOF (point and location) is held xed while the input DOF is moved, the only information that changes in Equation 8 as different FRFs are measured is the information relative to the modal coefcient for the particular mode of interest. Note that, since mechanical systems obey Maxwells Reciprocity Theorem (H pq = H qp ), either sensor (input or output) can be held xed with the other sensor allowed to rove or map the DOFs that dene the modal vector. {H( )} p Mr
2

{ r } pr + C r j + K r

(9)

If Equation 8 is evaluated (measured) near the undamped natural frequency, this means
+ -11Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

that the imaginary par t of the FRF will be proportional to the modal coefcient. The propor tionality constant r ) is:
r pr

M r 2 + C r j + K r

(10)

Since mode shapes are relative patterns, not absolute motions, the value of the constant is not important unless damping or modal scaling is required. Therefore, modal vectors can be estimated from the imaginary par t of the frequency response functions at the damped natural frequencies (or from the magnitude and phase information of the frequency response functions at the damped natural frequencies). This will be reasonably accurate as long as the undamped natural frequencies are well separated and the damping is small (undamped and damped natural frequencies nearly equal). This result is consistent with the expansion theorem concept (the response of the system at any instant in time or at any frequency is a linear combination of the modal vectors): Expansion Theorem - Time Domain: { x(t i ) } = Expansion Theorem - Frequency Domain: { X( i ) } =

r=1

r { r }

r=1

r { r }

Using the frequency domain form of the expansion theorem, if the response is evaluated at the undamped natural frequency of mode r, the expansion coefcient r will dominate and be approximately equal to alpha dened in Equation 9.

Damping Issues
Note that the experimental analysis assumes that damping is samll and constant for each mode, less than 10 percent of critical damping ( less than 0.10) and constant for all measurements. This is a practical assumption for many realistic, simple systems. Even with damping at 10 percent of critical damping, the difference between the damped natural frequency and the undamped natural frequency is less than one percent.

-12-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Appendix B: Impact Testing

-13-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Time Domain Histories

0.6

0.5

0.4

0.3

Volts
0.2 0.1 0 0.1 0

0.2

0.4

0.6 0.8 Time (Sec.)

1.2

1.4

Figure 11. Typical Force Signal (Impact) - Time Domain

0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 0.05 0.076 0.078 0.08 0.082 0.084 0.086 0.088 Time (Sec.) 0.09 0.092 0.094 0.096

Volts

Figure 12. Typical Force Signal (Impact) - Time Domain

-14-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Time Domain Histories

0.025 0.02 0.015 0.01 0.005

Volts

0 0.005 0.01 0.015 0.02 0.025

0.2

0.3

0.4

0.5

0.6 0.7 Time (Sec.)

0.8

0.9

1.1

Figure 13. Typical Force Signal (Impact) - Time Domain

0.06

0.04

0.02

Volts

0.02

0.04

0.06

0.2

0.4

0.6 0.8 Time (Sec.)

1.2

1.4

Figure 14. Typical Responce Signal (Exponential Decay) - Time Domain

-15-

Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

Frequency Domain Measurements

400

300

200

100

Phase (Degrees)

100

200

300

400

200

400

600

800 1000 Frequency (Hz)

1200

1400

1600

1800

Figure 15. Typical Frequency Response Function (Phase) - Frequency Domain

10
1

10

Magnitude (Volt/Volt)

10

10

10

200

400

600

800 1000 Frequency (Hz)

1200

1400

1600

1800

Figure 16. Typical Frequency Response Function (Magnitude) - Frequency Domain

1 0.9 0.8 0.7 0.6

Coherence

0.5 0.4 0.3 0.2 0.1 0

200

400

600

800 1000 Frequency (Hz)

1200

1400

1600

1800

Figure 17. Typical Coherence Function - Frequency Domain

+ -16Revision: September 24, 2003 +

+ 20-263-571

Structures/Motion Lab

UC-MINE +

-17-

Revision: September 24, 2003 +