Sie sind auf Seite 1von 24

1

DYNAMICS OF VIBRATING SYSTEMS


SUMMARY
This report has successfully demonstrated the analytical derivation of the governing equations for
the vibration of an unforced 6DoF damped mechanism, using Lagrange’s equations. MATLAB® was
successfully implemented in several instances. The system characteristics were found, namely the six
natural frequencies and modal shapes. A comprehensive modal analysis was performed with two co-
ordinate transformations and proportional damping. A particular use for MATLAB was performing
the impulse response for two separate non-periodic excitation forces applied to the system. Other
types of excitation were discussed, and finally some of the fundamental assumptions of the system
were altered to observe the effect on damping.

1. INTRODUCTION
In this report, the system in figure 1 with six Degrees of Freedom (DoF) will be considered. The
governing equation will be derived using Lagrangian mechanics in section 2. Then it will be subject to
a number of different analysis procedures. The matrix method will determine the undamped natural
frequencies and modal shapes in section 3. A kinematic analysis will be performed in section 4, with
modal analysis that uses two co-ordinate transformations and proportional damping. This process
will find estimates for damping ratios, damping factors and damped frequencies. A kinetic analysis
will take place in section 5, where two separate pre-determined non-periodic excitation forces will
be applied to the system. Numerical and analytical procedures are thereby examined for their
effectiveness in predicting the impulse response. Finally, in sections 6 and 7, alternative excitation
functions, and methods for modelling damping will be discussed.

k1 = 200 N/m k2 = 200 N/m k3 = 300 N/m k4 = 300 N/m k5 = 300 N/m k6 = 400 N/m k7 = 400 N/m

m1 = m2 = m3 = m4 = m5 = m6 =

10kg 20kg 30kg 20kg 40kg 60kg

c1 c2 c3 c4 c5 c6 c7
x1 x2 x3 x4 x5 x6
Figure 1 – 6DoF Mass-spring-damper model with parameters for mass and spring constants.

2
2. SYSTEM EQUATION
The analysis is presented for a purely translational system (where no rotation is involved). The
Lagrange equation for unforced oscillation with damping is given by:

𝑑 𝜕𝑇 𝜕𝑇 𝜕𝑉 𝜕𝐷
( ) − 𝜕𝑞 + 𝜕𝑞 + 𝜕𝑞̇ = 0
𝑑𝑡 𝜕𝑞̇
(2.1)
𝑖 𝑖 𝑖 𝑖

Where T is the Kinetic energy (J), V is the Potential energy (J), D is the energy dissipation due to
damping (J), qi is the co-ordinate system, 𝑞̇ 𝑖 is the first derivative of the co-ordinate qi with respect
to time (ms-1).

The Kinetic energy of the system given in figure 1 can be expressed

1 1 1 1 1 1
𝑇 = 2 𝑚1 𝑥̇ 1 2 + 2 𝑚2 𝑥̇ 2 2 + 2 𝑚3 𝑥̇ 3 2 + 2 𝑚4 𝑥̇ 4 2 + 2 𝑚5 𝑥̇ 5 2 + 2 𝑚6 𝑥̇ 6 2 (2.2)

Where mi is the respective mass for each of the masses (kg).

The potential energy can also be expressed thus:

1 1 1 1 1 1
𝑉 = 𝑘1 𝑥1 2 + 𝑘2 (𝑥2 − 𝑥1 )2 + 𝑘3 (𝑥3 − 𝑥2 )2 + 𝑘4 (𝑥4 − 𝑥3 )2 + 𝑘5 (𝑥5 − 𝑥4 )2 + 𝑘6 (𝑥6 − 𝑥5 )2 (2.3)
2 2 2 2 2 2

Where ki is the respective stiffness of each of the springs (Nm-1).

Energy dissipation due to damping is expressed:

1 1 1 1 1 1
𝐷 = 𝑐1 𝑥̇1 2 + 𝑐2 (𝑥̇ 2 − 𝑥̇1 )2 + 𝑐3 (𝑥̇ 3 − 𝑥̇ 2 )2 + 𝑐4 (𝑥̇ 4 − 𝑥̇ 3 )2 + 𝑐5 (𝑥̇ 5 − 𝑥̇ 4 )2 + 𝑐6 (𝑥̇ 6 − 𝑥̇ 5 )2 (2.4)
2 2 2 2 2 2

Where ci is the respective damping co-efficient for each damper.

By substituting into the Lagrange equation for i = 1,

𝜕𝑇 𝑑 𝜕𝑇 𝜕𝑉 𝜕𝐷
𝜕𝑞̇ 1
= 𝑚1 𝑥̇ 1 ( )
𝑑𝑡 𝜕𝑞̇ 1
= 𝑚1 𝑥̈ 1 𝜕𝑞1
= 𝑘1 𝑥1 − 𝑘2 (𝑥2 − 𝑥1 ) 𝜕𝑞̇ 1
= 𝑐1 𝑥̇ 1 − 𝑐2 (𝑥̇ 2 − 𝑥̇ 1 )

Using the Lagrange equation given in (2.1), the first equation (for i = 1) is therefore:

𝑚1 𝑥̈ 1 + (𝑘1 + 𝑘2 )𝑥1 − 𝑘2 𝑥2 + (𝑐1 + 𝑐2 )𝑥1 − 𝑐2 𝑥2 = 0 (2.5)

The same procedure may be continued for i = 2:

𝜕𝑇 𝑑 𝜕𝑇 𝜕𝑉
𝜕𝑞̇ 2
= 𝑚2 𝑥̇ 2 ( )
𝑑𝑡 𝜕𝑞̇ 2
= 𝑚2 𝑥̈ 2 𝜕𝑞2
= 𝑘2 (𝑥2 − 𝑥1 ) − 𝑘3 (𝑥3 − 𝑥2 )

𝜕𝐷
𝜕𝑞̇ 1
= 𝑐2 (𝑥̇ 2 − 𝑥̇ 1 ) − 𝑐3 (𝑥̇ 3 − 𝑥̇ 2 )

3
And hence the equation for i = 2 is given:

𝑚2 𝑥̈ 2 − 𝑘2 𝑥1 + (𝑘2 + 𝑘3 )𝑥2 − 𝑘3 𝑥3 − 𝑐2 𝑥1 + (𝑐2 + 𝑐3 )𝑥2 − 𝑐3 𝑥3 = 0 (2.6)

Since the system has 6 DoF, it follows the same pattern through to i = 6, the set of Lagrange
equations can now be presented in matrix format.

𝑚1 0 0 0 0 0 𝑥̈1
0 𝑚2 0 0 0 0 𝑥̈ 2
0 0 𝑚3 0 0 0 𝑥̈ 3
0 0 0 𝑚4 0 0 𝑥̈ 4
0 0 0 0 𝑚5 0 𝑥̈ 5
[0 0 0 0 0 𝑚6 ] {𝑥̈ 6 }
(𝑐1 + 𝑐2 ) −𝑐2 0 0 0 0 𝑥̇1
−𝑐2 (𝑐2 + 𝑐3 ) −𝑐3 0 0 0 𝑥̇ 2
0 −𝑐3 (𝑐3 + 𝑐4 ) −𝑐4 0 0 𝑥̇ 3
+
0 0 −𝑐4 (𝑐4 + 𝑐5 ) −𝑐5 0 𝑥̇ 4
0 0 0 −𝑐5 (𝑐5 + 𝑐6 ) −𝑐6 𝑥̇ 5
[ 0 0 0 0 −𝑐6 (𝑐6 + 𝑐7 )] {𝑥̇ 6 }
(𝑘1 + 𝑘2 ) −𝑘2 0 0 0 0 𝑥1
−𝑘2 (𝑘2 + 𝑘3 ) −𝑘3 0 0 0 𝑥2
0 −𝑘3 (𝑘3 + 𝑘4 ) −𝑘4 0 0 𝑥3
+ 𝑥4 = 0
0 0 −𝑘4 (𝑘4 + 𝑘5 ) −𝑘5 0
0 0 0 −𝑘5 (𝑘5 + 𝑘6 ) −𝑘6 𝑥5
[ 0 0 0 0 −𝑘6 (𝑘6 + 𝑘7 )] 𝑥6 }
{

(2.7)

By substituting the appropriate values from figure 1, Lagrange’s equations can be found for this
particular system.

10 0 0 0 0 0 𝑥̈1
0 20 0 0 0 0 𝑥̈ 2
0 0 30 0 0 0 𝑥̈ 3
0 0 0 20 0 0 𝑥̈ 4
0 0 0 0 40 0 𝑥̈ 5
[0 0 0 0 0 60] {𝑥̈ 6 }
(𝑐1 + 𝑐2 ) −𝑐2 0 0 0 0 𝑥̇1
−𝑐2 (𝑐2 + 𝑐3 ) −𝑐3 0 0 0 𝑥̇ 2
0 −𝑐3 (𝑐3 + 𝑐4 ) −𝑐4 0 0 𝑥̇ 3
+
0 0 −𝑐4 (𝑐4 + 𝑐5 ) −𝑐5 0 𝑥̇ 4
0 0 0 −𝑐5 (𝑐5 + 𝑐6 ) −𝑐6 𝑥̇ 5
[ 0 0 0 0 −𝑐6 (𝑐6 + 𝑐7 )] {𝑥̇ 6 }
400 −200 0 0 0 0 𝑥1
−200 500 −300 0 0 0 𝑥2
0 −300 600 −300 0 0 𝑥3
+ 𝑥4 = 0
0 0 −300 600 −300 0
0 0 0 −300 700 −400 𝑥5
[ 0 0 0 0 −400 800 ] {𝑥6 }

(2.8)

4
Equation (2.8) concludes the derivation of the governing equation for this particular system in matrix
form.

3. SYSTEM CHARACTERISTICS
In order to find the natural frequencies and modal shapes of the system, the Matrix method was
used chosen over the Holzer method because it is more appropriate for computational analysis.

([𝐴] − 𝜔𝑛 2 [𝐼])𝑥 = 0 (3.1)

Where;

[𝐴] = [𝑚]−1 [𝑘] (3.2)

Solving the characteristic equation for the eigenvalues of A, and then performing a square root on
the resultant matrix, gives the natural frequencies for the system. The MATLAB code in figure 2 was
used to do this:

K = [400,-200,0,0,0,0;-200,500,-300,0,0,0;0,-300,600,-300,0,0;0,0,-
300,600,-300,0;0,0,0,-300,700,-400;0,0,0,0,-400,800]
M =
[10,0,0,0,0,0;0,20,0,0,0,0;0,0,30,0,0,0;0,0,0,20,0,0;0,0,0,0,40,0;0,0,0,0,0
,60]
% Mass and stiffness matrices are set up.
A = inv(M)*K
% A is defined and then the eigenvalues & eigenvectors are found in order
% to solve according to the matrix method.
[v,d] = eig(A)
nat_freq = sqrt(d)
% A plot of all the modal shapes is formed
plot(v)
pause;
% Each time the user presses the return key, the modal shapes will appear
% in ascending order of natural frequency.
plot(v(:,5))
pause;
plot(v(:,6))
pause;
plot(v(:,4))
pause;
plot(v(:,3))
pause;
plot(v(:,2))
pause;
plot(v(:,1))
Figure 2 – MATLAB® code to solve for the natural frequencies and modal shapes

5
The code in figure 2 produced the following results for the natural frequencies.

7.1167 0 0 0 0 0
0 6.4541 0 0 0 0
0 0 5.0976 0 0 0
𝜔𝑛 =
0 0 0 4.2858 0 0
0 0 0 0 1.4514 0
[ 0 0 0 0 0 2.6591]

The modal shapes may be plotted from the corresponding eigenvectors. They are given on figure 3
combined in one plot, and they are singled out in ascending order from the lowest to highest
frequency in the subsequent figures 4 to 9, displayed on page 8.

7.1167 rad/s

6.4541 rad/s

1.4514 rad/s

2.6591 rad/s

4.2858 rad/s

5.0976 rad/s

FIGURE 3 - MODAL SHAPES. THE NATURAL FREQUENCIES ARE COLOUR CODED AS SHOWN ON THE LEFT OF THE FIGURE.

The dynamic characteristics of the system are the properties during its active and changing states.
Therefore discussing these involves noting the order in which the masses and stiffness elements for
that constitute each degree of freedom hit their natural frequency. The cause of this is due to the
surrounding properties of the system. This information is ordered and summarised in table 1.

Natural Frequency (ωn – [rad/s]) Mass (m – [kg]) Surrounding stiffness elements – Left to
right (k – [Nm-1])
1.4514 m5  40 k5  300; k6  400
2.6591 m6  60 k6  400; k7  400
4.2858 m4  20 k4  300; k5  300
5.0976 m3  30 k3  300; k4  300
6.4541 m2  20 k2  200; k3  300
7.1167 m1  10 k1  200; k2  200
TABLE 1 – NATURAL FREQUENCIES IN ASCENDING ORDER, ALLOCATED TO EACH MASS AND STIFFNESS ELEMENTS

6
The heavier mass and stiffer spring arrangements tend to have lower values for ωn. However, it may
be observed from table 1 that the lowest natural frequency actually occurs for m5, rather than m6.
Possible causes for this are the variance in surrounding stiffness elements for m5 whereas m6 has the
same k value on both sides; and also the fact that m5 has surrounding masses on both sides, whereas
m6 is against a wall on one side. Conversely, lighter masses with lower k values have higher natural
frequencies. This is a trend that loosely matches from m4 to m1. The exception occurs at m3, which is
located in the middle of the system, promoting a lower natural frequency. From the tabulation of
these results it is possible to see that the natural frequency for each of the six degrees of freedom is
highly dependent on every other element in the system.

7
FIGURE 4 – MODAL SHAPE FOR 1.4514 RAD/S FIGURE 5 – MODAL SHAPE FOR 2.6591 RAD/S

FIGURE 6 – MODAL SHAPE FOR 4.2858 RAD/S FIGURE 7 – MODAL SHAPE FOR 5.0976 RAD/S

FIGURE 8 – MODAL SHAPE FOR 6.4541 RAD/S FIGURE 9 – MODAL SHAPE FOR 7.1167 RAD/S

8
4. MODAL ANALYSIS
This is a kinematic analysis which approaches the 6DoF system assuming that free damped vibration
occurs. The damping will be proportional in order for the analysis to work, with α = 0.1 and β = 0.2.

The system equations of 2.8 were given thus:

10 0 0 0 0 0 𝑥̈1
0 20 0 0 0 0 𝑥̈ 2
0 0 30 0 0 0 𝑥̈ 3
0 0 0 20 0 0 𝑥̈ 4
0 0 0 0 40 0 𝑥̈ 5
[0 0 0 0 0 60] {𝑥̈ 6 }
(𝑐1 + 𝑐2 ) −𝑐2 0 0 0 0 𝑥̇1
−𝑐2 (𝑐2 + 𝑐3 ) −𝑐3 0 0 0 𝑥̇ 2
0 −𝑐3 (𝑐3 + 𝑐4 ) −𝑐4 0 0 𝑥̇ 3
+
0 0 −𝑐4 (𝑐4 + 𝑐5 ) −𝑐5 0 𝑥̇ 4
0 0 0 −𝑐5 (𝑐5 + 𝑐6 ) −𝑐6 𝑥̇ 5
[ 0 0 0 0 −𝑐 (𝑐 + 𝑐 )] {𝑥̇ 6 }
[ 6 6 7 ]
400 −200 0 0 0 0 𝑥1
−200 500 −300 0 0 0 𝑥2
0 −300 600 −300 0 0 𝑥3
+ 𝑥4 = 0
0 0 −300 600 −300 0
0 0 0 −300 700 −400 𝑥5
[ 0 0 0 0 −400 800 ] {𝑥6 }

(2.8 - repeated)

The problem was solved in MATLAB using the code given in figure 10.

9
% Stiffness and mass matrices defined, as before
K = [400,-200,0,0,0,0;-200,500,-300,0,0,0;0,-300,600,-300,0,0;0,0,-
300,600,-300,0;0,0,0,-300,700,-400;0,0,0,0,-400,800]
M =
[10,0,0,0,0,0;0,20,0,0,0,0;0,0,30,0,0,0;0,0,0,20,0,0;0,0,0,0,40,0;0,0,0,0,0
,60]
% Proportional damping applied to the system to form the damping matrix
C = 0.1*M+0.2*K

% Below - Used to check the order of undamped natural frequencies, will be


needed later for the P matrix.
A = inv(M)*K
[v,d] = eig(A)
nat_freq = sqrt(d)

% First co-ordinate transformation...


% Mass-Normalised Stiffness:
K_tilda = (M^-0.5)*(K)*(M^-0.5)

% Mass-Normalised Damping:
C_tilda = (M^-0.5)*(C)*(M^-0.5)

% Finding eigenvalues and eigenvectors for the normalised stiffness matrix


[v1,d1] = eig(K_tilda)

% Second co-ordinate transformation...


% Arranging the eigenvectors into ascending order according to natural
% frequency.
P = [v1(:,5),v1(:,6),v1(:,4),v1(:,3),v1(:,2),v1(:,1)]

% Diagonal matrices for use with the "r" co-ordinates after the second
% transformation
Big_Lambda_K = [P]'*[K_tilda]*[P]
Big_Lambda_C = [P]'*[C_tilda]*[P]

% Used for conversion from r(t) co-ordinates back to x(t) co-ordinates


[S] = M^-0.5*[P]

FIGURE 10 – MATLAB® CODE TO PERFORM KEY OPERATIONS REQUIRED FOR MODAL ANALYSIS.

The process of co-ordinate transformation and eventually finding the necessary parameters will be
detailed through each step.

Firstly, appropriate values should be substituted for C, for the damping matrix. The parameters α =
0.1 and β = 0.2 are given, and it is known that:

[𝐶] = 𝛼[𝑀] + 𝛽[𝐾]

Therefore:

𝑀{𝑥̈ } + (𝛼[𝑀] + 𝛽[𝐾]){𝑥̇ } + [𝐾]{𝑥} = 0

So, the equation of motion in terms of {x(t)} for 6DoF may be stated thus:

10
10 0 0 0 0 0 𝑥̈1 81 −40 0 0 0 0 𝑥̇1
0 20 0 0 0 0 𝑥̈ 2 −40 102 −60 0 0 0 𝑥̇ 2
0 0 30 0 0 0 𝑥̈ 3 0 −60 123 −60 0 0 𝑥̇ 3
+
0 0 0 20 0 0 𝑥̈ 4 0 0 −60 126 −60 0 𝑥̇ 4
0 0 0 0 40 0 𝑥̈ 5 0 0 0 −60 144 −80 𝑥̇ 5
[0 0 0 0 0 60] {𝑥̈ 6 } [ 0 0 0 0 −80 166 ] {𝑥̇ 6 }
400 −200 0 0 0 0 𝑥1
−200 500 −300 0 0 0 𝑥2
0 −300 600 −300 0 0 𝑥3
+ 𝑥4 = 0
0 0 −300 600 −300 0
0 0 0 −300 700 −400 𝑥5
[ 0 0 0 0 −400 800 ] {𝑥6 }

(4.1)

̃
In order to form a symmetric eigenvalue problem, the mass-normalised stiffness and damping, 𝐾
and 𝐶̃ , must be found:

1 1
̃ = [𝑀]−2 [𝐾][𝑀]−2
𝐾
1 1
𝐶̃ = [𝑀]−2 [𝐶][𝑀]−2

These matrices were calculated through the MATLAB calculations shown in figure 10:

400 −141.4214 0 0 0 0
−282.8427 500 −244.9490 0 0 0
̃= 0 −367.4235 600 −367.4235 0 0
𝐾
0 0 −244.9490 600 −212.1320 0
0 0 0 −424.2641 700 −326.5986
[ 0 0 0 0 −489.8979 800 ]

83 −29.6985 0 0 0 0
−59.3970 105 −51.4393 0 0 0
0 −77.1589 125 −75.9342 0 0
𝐶̃ =
0 0 −50.6228 126 −45.2548 0
0 0 0 −90.5097 150 −70.2187
[ 0 0 0 0 −105.3281 166 ]

̃ and 𝐶̃ can be expressed in terms of the new co-


To form the first co-ordinate transformation, 𝐾
ordinate system, {q(t)}.

̃ ]){𝑞̇ 𝑖 } + [𝐾
[𝐼]{𝑞̈ 𝑖 } + (0.1[𝐼] + 0.2[𝐾 ̃ ]{𝑞𝑖 } = 0

And therefore:

[𝐼]{𝑞̈ 𝑖 } + [𝐶̃ ]{𝑞̇ 𝑖 } + [𝐾


̃ ]{𝑞𝑖 } = 0

Substituting the mass-normalised and stiffness and damping matrices, equation (4.2) is formed:

11
1 0 0 0 0 0 𝑞̈ 1
0 1 0 0 0 0 𝑞̈ 2
0 0 1 0 0 0 𝑞̈ 3
0 0 0 1 0 0 𝑞̈ 4
0 0 0 0 1 0 𝑞̈ 5
[0 0 0 0 0 1] {𝑞̈ 6 }
83 −29.6985 0 0 0 0 𝑞̇ 1
−59.3970 105 −51.4393 0 0 0 𝑞̇ 2
0 −77.1589 125 −75.9342 0 0 𝑞̇ 3
+
0 0 −50.6228 126 −45.2548 0 𝑞̇ 4
0 0 0 −90.5097 150 −70.2187 𝑞̇ 5
[ 0 0 0 0 −105.3281 166 ] {𝑞̇ 6 }
400 −141.4214 0 0 0 0 𝑞1
−282.8427 500 −244.9490 0 0 0 𝑞2
0 −367.4235 600 −367.4235 0 0 𝑞3
+ 𝑞4 = 0
0 0 −244.9490 600 −212.1320 0
0 0 0 −424.2641 700 −326.5986 𝑞5
[ 0 0 0 0 −489.8979 800 ] {𝑞6 }

(4.2)

This completes the first co-ordinate transformation.

̃ were found with MATLAB. [P] is defined as the normalised


The eigenvalues and eigenvectors for 𝐾
matrix of stiffness eigenvectors, in ascending order of natural frequency.

0.1436 −0.1880 0.2826 0.4092 −0.3943 −0.7357


0.3848 −0.4379 0.4323 0.4055 0.0461 0.5539
0.5534 −0.4238 −0.0922 −0.5051 0.3925 −0.3104
𝑃=
0.4238 −0.0096 −0.4446 −0.1587 −0.7402 0.2228
0.4738 0.4687 −0.3811 0.5232 0.3601 −0.0754
[0.3457 0.6111 0.6180 −0.3377 −0.1038 0.0165 ]

𝛬𝐾̃ and 𝛬𝐶̃ will be used in the equation in terms of {r(t)}, the result of the second co-ordinate
transformation. They produce ordered values that relate to natural frequency and damping, as
shown below.

𝜔𝑛1 2 0 0 0 0 0
2
0 𝜔𝑛2 0 0 0 0
0 0 𝜔𝑛3 2 0 0 0
̃ ][𝑃] =
𝛬𝐾̃ = [𝑃]𝑇 [𝐾
0 0 0 𝜔𝑛4 2 0 0
0 0 0 0 𝜔𝑛5 2 0
[ 0 0 0 0 0 𝜔𝑛6 2 ]

12
2𝜁𝜔1 0 0 0 0 0
0 2𝜁𝜔2 0 0 0 0
0 0 2𝜁𝜔3 0 0 0
𝛬𝐶̃ = [𝑃]𝑇 [𝐶̃ ][𝑃] =
0 0 0 2𝜁𝜔4 0 0
0 0 0 0 2𝜁𝜔5 0
[ 0 0 0 0 0 2𝜁𝜔6 ]

The MATLAB calculations produced the following results for 𝛬𝐾̃ and 𝛬𝐶̃ :

2.1067 0 0 0 0 0
0 7.0710 0 0 0 0
0 0 18.3677 0 0 0
𝛬𝐾̃ =
0 0 0 25.9854 0 0
0 0 0 0 41.6552 0
[ 0 0 0 0 0 50.6474]

0.5213 0 0 0 0 0
0 1.5142 0 0 0 0
0 0 3.7735 0 0 0
𝛬𝐶̃ =
0 0 0 5.2971 0 0
0 0 0 0 8.4310 0
[ 0 0 0 0 0 10.2295]

To complete the second co-ordinate transformation, 𝛬𝐾̃ and 𝛬𝐶̃ are substituted into the system
equation expressed in terms of {r(t)} . This provides the format for decoupled equations to be
formed. Rather than expressing the system as a single 6DoF equation as before, it may now be
stated in terms of 6 separate single DoF (sDoF) equations.

[𝐼]{𝑟̈𝑖 } + [𝛬𝐶 ]{𝑟̇𝑖 } + [𝛬𝐾 ]{𝑟𝑖 } = 0

1 0 0 0 0 0 𝑟̈ 1 2𝜁𝜔1 0 0 0 0 0 𝑟̇ 1
0 1 0 0 0 0 𝑟̈ 2 0 2𝜁𝜔2 0 0 0 0 𝑟̇ 2
0 0 1 0 0 0 𝑟̈ 3 0 0 2𝜁𝜔 3 0 0 0 𝑟̇ 3
+
0 0 0 1 0 0 𝑟 ̈ 4 0 0 0 2𝜁𝜔 4 0 0 𝑟̇ 4
0 0 0 0 1 0 𝑟̈ 5 0 0 0 0 2𝜁𝜔5 0 𝑟̇ 5
[0 0 0 0 0 1] {𝑟̈ 6 } [ 0 0 0 0 0 2𝜁𝜔6 ] {𝑟̇ 6 }
𝜔𝑛1 2 0 0 0 0 0
2
𝑟1
0 𝜔𝑛2 0 0 0 0 𝑟2
0 0 𝜔𝑛3 2 0 0 0 𝑟3
+ 2 𝑟4 = 0
0 0 0 𝜔𝑛4 0 0
𝑟5
0 0 0 0 𝜔𝑛5 2 0
{𝑟6 }
[ 0 0 0 0 0 𝜔𝑛6 2 ]

13
Substituting in values for 𝛬𝐾̃ and 𝛬𝐶̃ gives (4.3):

1 0 0 0 0 0 𝑟̈ 1 0.5213 0 0 0 0 0 𝑟̇ 1
0 1 0 0 0 0 𝑟̈ 2 0 1.5142 0 0 0 0 𝑟̇ 2
0 0 1 0 0 0 𝑟̈ 3 0 0 3.7735 0 0 0 𝑟̇ 3
+
0 0 0 1 0 0 𝑟̈ 4 0 0 0 5.2971 0 0 𝑟̇ 4
0 0 0 0 1 0 𝑟̈ 5 0 0 0 0 8.4310 0 𝑟̇ 5
[0 0 0 0 0 ]
1 {𝑟̈ 6 } [ 0 0 0 0 0 ]
10.2295 {𝑟̇ 6 }
2.1067 0 0 0 0 0 𝑟1
0 7.0710 0 0 0 0 𝑟2
0 0 18.3677 0 0 0 𝑟3
+ 𝑟4 = 0
0 0 0 25.9854 0 0
0 0 0 0 41.6552 0 𝑟5
[ 0 0 0 0 0 50.6474] {𝑟6 }

(4.3)

The equivalent 6 sDoF equations are:

𝑟̈1 + 0.5213𝑟̇1 + 2.1067𝑟1 = 0

𝑟̈2 + 1.5142𝑟̇2 + 7.0710𝑟2 = 0

𝑟̈3 + 3.7735𝑟̇3 + 18.3677𝑟3 = 0

𝑟̈4 + 5.2971𝑟̇4 + 25.9854𝑟4 = 0

𝑟̈5 + 8.4310𝑟̇5 + 41.6552𝑟5 = 0

𝑟̈6 + 10.2295𝑟̇6 + 50.6474𝑟6 = 0

To obtain the final solution for x{t}, simply:


1
{𝑥(𝑡)} = [𝑀]−2 [𝑃]{𝑟(𝑡)} = [𝑆]{𝑟(𝑡)} (4.4)

Where:
1
[𝑆] = [𝑀]−2 [𝑃]

14
In order to tabulate the results for each DoF, each of the sDoF equations found in terms of {r(t)}
must be addressed (equation (4.3)). Therefore the natural frequencies are calculated from the
values of 𝛬𝐾̃ , where each term is equal to 𝜔𝑛𝑖 2 :

𝜔𝑛1 = √2.1067 = 1.4514 𝑟𝑎𝑑/𝑠

𝜔𝑛2 = √7.071 = 2.6591 𝑟𝑎𝑑/𝑠

𝜔𝑛3 = √18.3677 = 4.2858 𝑟𝑎𝑑/𝑠

𝜔𝑛4 = √25.9854 = 5.0976 𝑟𝑎𝑑/𝑠

𝜔𝑛5 = √41.6552 = 6.4541 𝑟𝑎𝑑/𝑠

𝜔𝑛6 = √50.6474 = 7.1167 𝑟𝑎𝑑/𝑠

The damping factors are found directly from each value of 𝛬𝐶̃ . The damping ratios are found from
the each value of 𝛬𝐶̃ where each term is equal to 2𝜁𝜔𝑖 :

0.5213
𝜁1 = = 0.1796 𝑁𝑠/𝑚
(2 ∗ √2.1067)

1.5142
𝜁2 = = 0.2847 𝑁𝑠/𝑚
(2 ∗ √7.071)

3.7735
𝜁3 = = 0.4402 𝑁𝑠/𝑚
(2 ∗ √18.3677)

5.2971
𝜁4 = = 0.5196 𝑁𝑠/𝑚
(2 ∗ √25.9854)

8.4310
𝜁5 = = 0.6532 𝑁𝑠/𝑚
(2 ∗ √41.6552)

10.2295
𝜁6 = = 0.7187 𝑁𝑠/𝑚
(2 ∗ √50.6474)

Finally, damped frequencies (𝜔𝑑 ) are found from equation (4.5).

𝜔𝑑 = 𝜔𝑛 √1 − 𝜁𝑖 2 (4.5)

15
Natural frequencies, 𝜔n Damping ratio, ζ Damping Factor Damped frequencies, 𝜔d
(rad/s) (Ns/m) (Ns/m) (rad/s)
1.4514 0.1796 0.5213 1.4279
2.6591 0.2847 1.5142 2.5491
4.2858 0.4402 3.7735 3.8481
5.0976 0.5196 5.2971 4.3555
6.4541 0.6532 8.4310 4.8872
7.1167 0.7187 10.2295 4.9484
TABLE 2: COMPARISON BETWEEN PARAMETERS OBTAINED THROUGH MODAL ANALYSIS.

Table 2 shows the trend between parameters obtained through

From table 2 it may be observed that damping ratio increases with the natural frequency. This is
intuitive because with a higher oscillating frequency it is likely that more damping occurs. The
damped frequencies increase with natural frequency, but not at a linear rate, as shown by graph 1.
The relationship between them is actually convergent because damping factor increases to a limit of
1. Therefore at higher values the damped frequencies are significantly lower than the natural
frequencies.

6
Frequency (rad/s)

4
Natural Frequencies
3 Damped Frequencies

0
0.0 0.2 0.4 0.6 0.8
Damping ratio (Nm/s)

GRAPH 1: COMPARISON BETWEEN NATURAL FREQUENCIES AND DAMPED FREQUENCIES ACROSS DAMPING RATIOS.

16
5. NON-PERIODIC EXCITATION

5.1 IMPULSE FORCE (FORCE_1.M)


This section performs a kinetic analysis on the system shown in figure 1 for an impulse force acting
specifically on m1. The impulse force is defined as 𝐹̂ .

𝐹̂ = ∫ 𝐹(𝑡) 𝑑𝑡 (5.1)

When considering a unit impulse:

𝐹̂ = 𝑓(𝜉)∆ 𝜉 (5.2)

The convolution integral can be formed:


𝑡
𝑥(𝑡) = ∫0 𝑓(𝜉)ℎ(𝑡 − 𝜉)𝑑𝜉 (5.3)

The rectangular pulse can be considered analytically as the sum of two step functions, where F(t) =
F0 for a damped system.
𝐹0
𝑥1 (𝑡) = 𝑘
(1 − 𝑐𝑜𝑠𝜔𝑛 𝑡) 𝑡 < 𝑡1 (5.4)

𝐹0
𝑥2 (𝑡) = 𝑘
[1 − 𝑐𝑜𝑠𝜔𝑛 (𝑡 − 𝑡1 )] 𝑡 = 𝑡1 (5.5)

The sum of which is:


𝐹0
𝑥(𝑡) = 𝑘
[−𝑐𝑜𝑠𝜔𝑛 𝑡 + 𝑐𝑜𝑠𝜔𝑛 (𝑡 − 𝑡1 )] 𝑡 > 𝑡1 (5.6)

For this system, the values are:


1
𝑥(𝑡) = 200 [−cos(7.1167𝑡) + cos(7.1167(𝑡 − 𝑡1 ))] (5.7)

The original excitation files for use with MATLAB are given in Appendix 2. To apply these forces to
the particular system shown in figure 1, additional sections of code were added to perform the
transfer function, impulse response and convolution integral. In figures 11A and 12A, the original
sections are shaded out in grey, and the added sections are in black.

The parameters for the transfer function were formed thus:

1
′ (𝑠)
𝑋(𝑠) 𝑚
𝐺 = =
𝐹(𝑠) 𝑠 2 + 2𝜁𝜔𝑛 𝑠 + 𝜔𝑛 2
1
For m1: 𝑚 = 0.1, 2𝜁𝜔𝑛6 = 10.2295 and 𝜔𝑛6 2 = 50.6474

The impulse solution is given on the next page, using a linear simulation and convolution integral.
The plots are given in order of the corresponding solutions noted by the code.

17
dt=0.1;
t=0:dt:50;
f=[zeros(50,1);ones(100,1);zeros(length(t)-150,1)];
subplot(311);
plot(t,f);
title('Square input');
axis([0 50 0 1.5]);
pause;

%Numerical solution – impulse response


sys=tf(0.1,[1 10.2295 50.6474]);% using m1 = 10kg and values from sect 4
% in the stiffness and damping matrices in terms of r(t) for nat freq #6
h=impulse(sys,t);
sol_1=lsim(sys,f,t); %linear simulation
sol_2=conv(f,h)*dt; %convolution
subplot(312);
plot(t,sol_1(1:length(t)),t,sol_2(1:length(t)));
title('Impulse response of system');
xlabel('Time');
ylabel('Output');
pause;

%Attempted analytical validation – step response


y=0.005.*((-cos(7.1167*t))+(cos(7.1167.*(t-(t-0.1))))) %Using Nat. Freq #6.
subplot(313);
plot(t,y)
title('System Response to a step input');
xlabel('Time');
ylabel('x(t)');
FIGURE 11A – MATLAB® CODE FOR THE NON-PERIODIC EXCITATION FORCE_1.M APPLIED TO m1 IN THE 6DOF SYSTEM.

FIGURE 11B – IMPULSE RESPONSE PLOT FOR FORCE_1.M APPLIED TO m1 IN THE 6DOF SYSTEM
18
The impulse response showed a resemblance to the rectangular input used to excite m1. The system
reacts dynamically for the duration of the impulse, and then the force dies away over time due to
the system damping. The analytical result for step input did not produce a result as expected as it is
a continuous function rather than a solution to the step input. This is due to an unknown flaw in the
equation presented in the code.

5.2 HALF-SINE PULSE EXCITATION FORCE (FORCE_2.M)


In this case a half-sine pulse was applied to m1 on the system given in figure 1.

The analytical solution in this case is given by:

𝐹0 𝑒 −𝜁𝜔𝑛𝑡
𝑥(𝑡) = [1 − ] sin(𝜔𝑑 𝑡 − 𝜙) (5.8)
𝑘 √1−𝜁 2

Where:

𝜁
𝜙 = 𝑡𝑎𝑛−1
√1 − 𝜁 2

Stating (5.8) with values relevant to this problem:

1 𝑒 −5.1148𝑡
𝑥(𝑡) = 200 [1 − ] sin(4.9484𝑡 − 0.802) (5.9)
√1−0.71872

The solution using MATLAB code for a linear simulation and convolution integral is given on the
following page.

The plots are given in order of the corresponding solutions noted by the code.

19
DT=0.1;
t=0:DT:50; %Time (high resolution)
sys=tf(0.1,[1 0.5213 2.1067]);
f=(sin(t).*t+0.1).*(t<8.1);
subplot(311);
plot(t,f,'b');
xlabel('Time');
ylabel('Input');
title('Input vs Time');
pause;

% Impulse response (below)


h=impulse(sys,t);
sol_1=lsim(sys,f,t); %linear simulation
sol_2=conv(f,h)*dt; %convolution
subplot(312);
plot(t,sol_1(1:length(t)),t,sol_2(1:length(t)))
title('Impulse response of system');
xlabel('Time');
ylabel('Output');
pause;

% Attempted Analytical validation - step response (below)


y = 0.2.*(1-(exp(-5.1148*t)/sqrt(1-0.7187^2).*(sin(4.9484*t-
(atan(0.7187/sqrt(1-0.7187^2)))))))
subplot(313);
plot(t,y)
title('System Response to a step input');
xlabel('Time');
ylabel('x(t)');
axis([0 5 0 0.5]);
FIGURE 12A – MATLAB® CODE FOR THE NON-PERIODIC EXCITATION FORCE_2.M APPLIED TO m1 IN THE 6DOF SYSTEM

FIGURE 12B – RESPONSE PLOT FOR FORCE_2.M APPLIED TO m1 IN THE 6DOF SYSTEM
20
The system response appears to be valid for the convolution integral approach (impulse response,
middle plot). Here, the vibrations follow the trend of the input before dying away due to damping in
the system. This shows that a numerical approach can provide accurate results with appropriate
conditions for linear simulation. Unfortunately the validation was not very accurate for the half-sine
pulse (analytical approach, lower plot). This could be due to an error in the equation or a mistake
with the input for MATLAB. However it still shows an initial transient period followed by stability.

6. OTHER TYPE OF EXCITATION


An excitation is considered of the following form in (6.1):

𝐹(𝑡) = 𝑒 −𝑡 𝑠𝑖𝑛𝜔𝑡 (6.1)

To illustrate the implications of this on the system, the function is plotted for two different values of
ω, the oscillating frequency, in figures 13 and 14.

FIGURE 13 – FORCING FUNCTION (6.1) APPLIED FIGURE 14 – FORCING FUNCTION (6.1) APPLIED
WHERE ω = 5 RAD/S WHERE ω = 50 RAD/S

Therefore it is a function which dies away in amplitude at a fixed rate, regardless of the natural
frequency. It is clear to see that this is unlike the other excitation functions in section 5 which were
largely solved through numerical analysis, as it is dissimilar to simple rectangular, half-sine pulse or
constant with rise time functions.

The best way to tackle this problem analytically is through the forced modal analysis method.

[𝑀]{𝑥̈ } + [𝐶]{𝑥̇ } + [𝐾]{𝑥} = [𝐵]{𝐹(𝑡)} = [𝐵]{𝑒 −𝑡 𝑠𝑖𝑛𝜔𝑡} (6.2)

Proportional damping would occur in the same way as in section 4, and the co-ordinate
transformations would also be applied to the right hand side of the equation. [B] is defined here as
an identity matrix.

21
7. EFFECTS OF DAMPING

7.1 SELECTING PROPORTIONAL DAMPING CONSTANTS


In section 4 it was assumed that the values for α and β we given. But this raises the question as to
how they were found in the first place. When the system is in terms of r co-ordinates, its equations
follow this format for unforced vibration:

{𝑟̈ (𝑡)} + (𝛼[𝐼] + 𝛽[𝛬]){𝑟̇ (𝑡)} + [𝛬]{𝑟(𝑡)} = 0 (7.1)

The damping constant, c, can be found through empirical measurements or chosen for desired
values.

𝑐
𝜁𝑖 = 2√𝑘𝑚 (7.2)

Then, through the relationship given in (7.2), The damping ratios, ζi , can thus be found. Proportional
values of α and β can be chosen such that these values of ζi are matched.

7.2 SYSTEM RESPONSE WITHOUT HORIZONTAL ROLLERS


If the masses were resting on a flat surface this would introduce additional losses through friction.
Due to the conservation of energy, the governing equation would follow this form:

[𝑀]{𝑥̈ } + [𝐶]{𝑥̇ } + [𝐾]{𝑥} + ∑ 𝐹𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝐿𝑜𝑠𝑠𝑒𝑠 = 0

The precise values for frictional losses could either be measured empirically or from the time taken
for the vibrations of the system to diminish. The latter method assumes that all other damping
parameters in the system are known.

7.3 NON-RIGID WALLS


Rigid walls essentially act as a place for the energy to “bounce” back through the system. This goes
by the assumption that no energy is lost at the walls and that it is all transferred back into the
system. If the walls lost their rigidity, there would have to be a proportional constant for how much
of the energy was lost each time a vibration hit a wall and travelled back into the system.

22
8. CONCLUSIONS
The 6DoF system was treated in a number of ways, both computationally and analytically. The
system equations and characteristics were successfully found. Modal analysis provided a useful way
of tackling vibration problem in conjunction with MATLAB, which computed the co-ordinate
transformations. All of the natural frequencies, damping ratios, damping factors and damped
frequencies were found. Forced, non-periodic excitation was analysed numerically with the
convolution integral, for a rectangular pulse and a half-sine pulse. The results were accurate using
the impulse response in MATLAB. However, the analytical validation proved to be a challenge and
the results were likely to be incorrect. Forced modal analysis was considered to model an excitation
function of 𝐹(𝑡) = 𝑒 −𝑡 𝑠𝑖𝑛𝜔𝑡. Overall, the analysis of this 6DoF system proved to be successful and a
good understanding has been formed of its fundamental properties in free and forced vibration.

23
9. REFERENCES
[1] TONGUE, B., 1996, Principles of Vibration. Oxford University Press

[2] THOMSON W. T., DAHLEH M. D., 1998, Theory of Vibration with Applications, 5th ed. Prentice Hall.

[3] ES386 Student Resources. Available online:


http://www2.warwick.ac.uk/fac/sci/eng/euo/modules/year3/es386a/resources/

10. APPENDICES

APPENDIX 1
ES386 Assignment handout: Dr. Xian Ping Liu, School of Engineering, Warwick University. March
2010.

APPENDIX 2
Excitation files for sections 5.1 and 5.2

Force_1.m

dt=0.1;
t=0:dt:50;
f=[zeros(50,1);ones(100,1);zeros(length(t)-150,1)];
subplot(211)
plot(t,f)
title('Square input')
axis([0 50 0 1.5])
%pause

Force_2.m

DT=0.1;
t=0:DT:50; %Time (high resolution)
f=(sin(t).*t+0.1).*(t<8.1);
plot(t,f,'b');
xlabel('Time');
ylabel('Input');
title('Input vs Time');
pause;

24

Das könnte Ihnen auch gefallen