You are on page 1of 10

Multi-degree-of-freedom System Frequency Response Function Curve-fitting

By Tom Irvine
Email: tom@vibrationdata.com
January 22, 2014
______________________________________________________________________________

Variables

Excitation frequency

fr

Natural frequency for mode r

Total degrees-of-freedom

H i j (f )

The steady state displacement at coordinate i due to a harmonic force excitation


only at coordinate j

Damping ratio for mode r

i r

Mass-normalized eigenvector for physical coordinate i and mode number r

Excitation frequency (rad/sec)

Natural frequency (rad/sec) for mode r

Receptance
The steady-state displacement at coordinate i due to a harmonic force excitation only at coordinate j is

1
ir j r
H i j (f )

2
2
r 1 r
1 r j 2 r r

(1)

where

r f / f r

(2)

j 1

(3)

Mobility
The steady-state velocity at coordinate i due to a harmonic force excitation only at coordinate j is

1
ir j r
(f ) j
H

ij
2
2
r 1 r
1 r j 2 r r

(4)

Accelerance
The steady-state acceleration at coordinate i due to a harmonic force excitation only at coordinate
j is

1
i
r
j
r
~
H i j (f ) 2

2
2
r 1 r
1 r j 2 r r

(5)

Curve-fitting Equation
Consider that a measured receptance function is available. Estimate the number of modes N within a
selected frequency band. Form a trial function H trial(f ).

N A

1
r
H trial(f )
2

r 1 r 1 r j 2 r r

(6)

Generate a set of trial functions by randomly varying the amplitudes A r , natural frequencies r , and
modal damping ratios r . Initial bounds and estimates may be set for each of these parameters. Subtract
each trial function from the measured data to determine which one yields the least residual error. The
final chosen function will then yield the modal parameters.

References
1.

T. Irvine, An Introduction to Frequency Response Functions, Vibrationdata, 2000.

2.

T. Irvine, Calculating Transfer Functions from Normal Modes, Revision E, Vibrationdata, 2013.

APPENDIX A

Example

x3
m3

k3
k5

x2
m2

k4

k2
x1
m1

k1

Figure A-1.

A three-degree-of-freedom system is shown in Figure A-1. First, determine the displacement using the
full mode set. Then solve for two modes only. Finally solve using mode acceleration with two modes via
equation (12). Compare the results at mass 3.

The parameters are


m1

0.0895

lbf sec^2/in

m2

0.0887

lbf sec^2/in

m3

0.0770

lbf sec^2/in

k1

1.8522e+04

lbf/in

k2

0.2157e+04

lbf/in

k3

0.2270e+04

lbf/in

k4

1.9429e+04

lbf/in

k5

1.7072e+04

lbf/in

The damping is 0.05 for all modes.

The mass matrix is

0
0.0895
0
0.0887

0
0

0 lbf sec 2 / in
0.0770
0

(A-1)

The stiffness matrix is

0
20679 - 2157
- 2157 23856 - 2270 lbf / in

0
- 2270 19342

(A-2)

The natural frequencies are

73.639

78.277 86.476 Hz

(A-3)

The mode shapes are

2.5194 - 1.9462 1.0190


1.7737 0.8756 - 2.7131

1.4090 2.7751 1.8167

(A-4)

Figure A-2.

The curve-fit will be performed on the H33 frequency response function.


The mass and stiffness parameters were chosen so that the modal frequencies would be closely-spaced as
a rigorous test of resolution abilities of the curve-fit method.

Figure A-3.

The H33 complex receptance FRF is shown in Figure A-3.

Figure A-4.

Three natural frequencies are estimated from the close-up view of the complex receptance function.

Figure A-5.

The curve-fitting is performed using Matlab script: mdof_frf_curvefit.m. Excellent agreement was
obtained.
A comparison of the analytic values and the numerical experiment curve-fit values are shown in the
following tables.

Table A-1. FRF Curve-Fit


Mode

fn(Hz)

Damping Ratio

73.29

0.0452

78.18

0.0526

86.55

0.0488

Table A-2. Natural Frequency Comparison


Mode

Analysis
fn(Hz)

FRF Curve-Fit
fn(Hz)

Difference

73.6394

73.29

0.5%

78.2766

78.18

0.1%

86.4757

86.55

-0.1%

Table A-3. Damping Comparison


Mode

Analysis
Damping Ratio

FRF Curve-Fit
Damping Ratio

Difference

0.05

0.0452

9.6%

0.05

0.0526

-5.2%

0.05

0.0488

2.4%

10

You might also like