Curitiba
2005
DEDICATRIA
iii
AGRADECIMENTOS
iv
SUMRIO
v
6 RESULTADOS.................................................................................................. 93
7 CONCLUSO................................................................................................... 94
8 SUGESTO DE TRABALHO FUTURO...................................................... 95
REFERNCIAS BIBLIOGRFICAS............................................................... 96
ANEXO 1............................................................................................................... 102
ANEXO 2............................................................................................................... 113
vi
LISTA DE FIGURAS
vii
LISTA DE TABELAS
viii
LISTA DE SMBOLOS
ix
RESUMO
x
ABSTRACT
xi
1 INTRODUO
d T T U
+ = Fqi (1)
dt q& i q i qi
associada a q i .
Utilizando o mtodo de elementos finitos possvel descrever a dinmica do rotor.
As equaes correspondentes dos elementos se apresentam de forma matricial.
6
2.1 O DISCO
( )
M D = R2 r 2 h D (2)
I Dx 0 0
ID = 0 I Dy 0 (3)
0 0 I Dz
MD
(
I Dx = I Dz = 12 3r + 3R + h
2 2 2
)
onde,
M
(
I Dy = D r 2 + R 2
2
)
7
r i = Ri + Ai u i (4)
onde:
Ri - o vetor posio global da origem do sistema de coordenada fixada no corpo
(xi yizi);
Ai - a matriz de transformao que define a orientao do corpo no sistema de
coordenada global (XYZ);
u i - o vetor posio de um ponto arbitrrio com relao origem do sistema de
coordenada fixo no corpo.
8
Ai uma matriz 3 x 3
R i = [ R xi R iy Rzi ] T
[
u i = u xi u yi u zi ] = [x
T i
yi zi ] T
cos sen 0
A = sen cos 0 (5)
0 0 1
1 0 0
A = 0 cos sen (6)
0 sen cos
cos 0 sen
A = 0 1 0 (7)
sen 0 cos
9
Assim, a orientao final do disco (coordenada xi yi zi) pode ser obtida no sistema
XYZ pela seguinte matriz de transformao:
Ai = A . A . A (7)
cos cos sen sen sen sen cos cos sen + sen sen cos
A = sen cos + sen sen cos
i
cos cos sen sen cos sen cos (8)
cos sen sen cos cos
ou R 0 ( X , Y , Z ) = Ai R ( x I , y i , z i ) .
T
v r
Considerando que K o vetor unitrio ao longo do eixo Z, i1 o vetor unitrio ao
v
longo eixo x1 e j o vetor unitrio ao longo do eixo y. Escrevendo as componentes dos
r v v
trs versores nas direes i , j e K do sistema de referncia R(xiyizi) tem-se as
seguintes relaes:
0
v
j = 1 (9)
0
T DR/ R =
2
[
1 vR
R / R0 ] [I ][v
T
D
R
R / R0 ] (15)
I Dx 0 0 x
T R
D /R
1
= x
2
[ y z ] 0
I Dy 0 y (16)
0 0 I Dz z
TDR/ R =
1
(
I Dx x2 + I Dy y2 + I Dz z2 ) (17)
2
TOR/0 R0 =
1
(
M D u& 2 + w& 2 ) (18)
2
T D = TOR/0 R0 + T DR/ R
TD =
1
( 1
) (
M D u& 2 + w& 2 + I Dx x2 + I Dy y2 + I Dz z2 ) (19)
2 2
No caso particular que est sendo trabalhado, o disco simtrico, verificando-se
que I Dx = I Dz . Por outro lado, supe-se que a velocidade angul ar permanece constante,
cos 1
(20)
sen
1 1 1 1
TDR/ R = I Dx [( 2x + z2 )] + I Dy2y = I Dx [(& sen + & cos )2 + (& cos & sen )2 ] + I Dy (& +)2
2 2 2 2
1
TDR/ R = [ I Dx (& 2 +& 2 ) + I Dy ( 2 + 2& +& 2 2 )] (22)
2
TD =
1
( 1
) ( 1
) (
M D u& 2 + w& 2 + I Dx & 2 + & 2 + I Dy 2 + 2& + & 2 2 ) (23)
2 2 2
TD
1
( 1
) ( 1
) (
M D u& 2 + w& 2 + I Dx & 2 + & 2 + I Dy 2 + 2& ) (24)
2 2 2
1
Na expresso anterior , pode-se observar que o termo I Dy 2 uma constante, no
2
tem nenhuma influncia nas equaes de Lagrange e representa a energia do disco que
gira a uma velocidade . O ltimo termo I Dy & representa o efeito giroscpico.
13
Como hiptese simplificativa, considera-se neste trabalho que o disco possui uma
rigidez infinita, portanto suas deformaes so nulas. A contribuio da energia
potencial ao disco nula, estando o mesmo caracterizado somente pela energia
cintica.
q i = [u i , wi , i , i ]
T
(25)
d TD TD
= Fq i (26)
dt q& i q i
d TD TD d 1
=
dt q& i qi dt q& i 2
M D
&
u(i
2
+ &
w i)2
+
1
2
I (
Dx i )
& 2 +& 2 + 1 I 2 + 2&
i
2
(
Dy )
i i
1
qi 2
( 2
)2 1
( 2
) 2 1
( )
M D u& i + w& i + I Dx &i +& i + I Dy 2 + 2& i i =
(27)
2 2
= M D qi + GD qi
&& &
14
M D 0 0 0
0 MD 0 0
Md = (28)
0 0 I Dx 0
0 0 0 I Dx
Utilizando os comentrios feitos para a matriz de inrcia, o disco contribui com a matriz
giroscpica com a expresso:
0 0 0 0
0 0 0 0
G D = (29)
0 0 0 I Dy
0 0 I Dy 0
logo:
M D 0 0 0 u&&i 0 0 0 0 u& i
w 0 w& i
d TD TD 0 MD 0 0 && i + 0 0 0
= (30)
dt q& i q i 0 0 I Dx 0 &&i 0 0 0 I Dy &i
0 0 0 I Dx && i 0 0 I Dy 0 & i
15
2.2 O EIXO
O eixo foi modelado como uma viga de seo transversal circular S = d2/4 e
momento de segunda ordem de rea I = d4/64, e caracterizado pela energia
potencial e a energia cintica.Tambm se considera que o material do eixo possui uma
densidade constante , mdulo de elasticidade E e um coeficiente de Poisson .
(& )
S I
(u& )
L L L
TE = 2
+ w& 2 dy + 2
+ & 2 dy + IL 2 + 2 I & dy (31)
2 0
2 0 0
2 2
2u * 2 w * 1 u * 1 w *
= x z + + (32)
y 2 y 2 2 y 2 y
ou
= l + nl (33)
1
UE =
2 T d (34)
17
= E (35)
tem-se:
1
2
UE = ( l + nl ) E( l + nl ) d (36)
UE =
E
2 ( )
l2 + 2 l nl + nl2 d (37)
nl l d = 0 (38)
Por outro lado, o terceiro termo dentro da integral na equao (37) de segunda
ordem e ser desprezado. Logo, a expresso para calcular a energia potencial dada
por:
L 2
x u z w
2 * 2 *
E dS dy
UE = (39)
2 y 2 y 2
0 S
18
E L 2 2 u* 2u * 2 w*
2 2 * 2
2 w
U E = x 2 + z + 2 xz dS dy (40)
2 0 S y y 2 y 2 y 2
Por causa da simetria que possui a seco transversal do eixo, neste trabalho sero
estudados eixos simtricos, sendo o terceiro termo da expresso integral anterior nulo.
I x = z 2 dS (41)
S
I z = x 2 dS (42)
S
E L 2u *
2 2
2w*
U E = 2 x dS + 2 z 2dS dy
2
2 0 y S y S
E L 2u *
2 2
2 w*
U E = I z 2 + I x 2 dy (43)
2 0 y y
u * = u cos t wsen t
(44)
w* = usen t + w cos t
E 2w
2 2
2u 2w 2u
L
EI 2 u 2 w
L 2 2
2 0 y 2 y 2
UE = + dy (46)
w
= (47)
y
u
= (48)
y
q e = [u i , wi , i , i , u i +1 , wi +1 , 1+ i , i+1 ]
T
(49)
u = [u i , i , u i +1 , i +1 ]
T
(50)
w = [wi , i , wi+1 , i +1 ]
T
(51)
21
u = N1 ( y )u (52)
w = N 2 ( y ) w (53)
onde N1(y) e N2(y) so funes de interpolao. Para este caso particular, foram
adotadas as funes de deslocamento tpico de uma viga apoiada (Lalanne e Ferraris,
1990):
3y 2 2 y 3 2y 2 y 3 3 y 2 2 y 3 y 2 y3
N 2 ( y ) = 1 2 + 3 ; y + 2 ; 2 3 ; +
L L2
(55)
L L L L L L
[u& ]
S L
2 0
I L T dN1 T dN 1 dN 2
T
T dN 2
2 0
+ &
u &
u + &
w w& dy (56)
dy dy dy dy
L
dN1 T dN 2
2 I u& T w dy + IL 2
0
dy dy
S T L T
L
TE = u N1 N 1 dy u + w N 2 N 2 dy w& +
& & & T T
2 0 0
I T dN1 dN 1
L T L T
dN 2 dN 2
+ u& dy u& + w& T dy w& (57)
2 0
dy dy 0
dy dy
L T
dN 1 dN 2
2 Iu& dy w + IL 2
T
0
dy dy
A equao anterior pode ser simplificada. O produto N1T N1 forma uma matriz de
1 T 1 1 1
TE = u& M 1u& + w& T M 2w& + u& T M 3u& + w& T M 4 w& + u& T M 5 w + IL 2
2 2 2 2
(58)
Onde:
d TE TE
e = (M Ec + M Eg )q&& e + Gq& e (59)
dt q& e q
onde q e o vetor de deslocamento do elemento dado pela equao (49), MEc surge de
M1 e M2, MEg de M3 e M4 e G de M5. A seguir so descritas estas trs matrizes:
156 0 0 22 L 54 0 0 13 L
0 156 22 L 0 0 54 13 L 0
0 22 L 4 L2 0 0 13 L 3L 2 0
SL 22 L 13 L 3L2
2
0 0 4L 0 0
M Ec = (60)
420 54 0 0 13 L 156 0 0 22 L
0 54 13 L 0 0 156 22 L 0
0 13 L 3L2 0 0 22 L 4L 2 0
13 L 0 0 3L2 22 L 0 0 4 L2
24
36 0 0 3 L 36 0 0 3L
0 36 3L 0 0 36 3L 0
0 3L 4L2 0 0 3L L2 0
I 3L L2
2
0 0 4L 3L 0 0
M Eg = (61)
30 L 36 0 0 3L 36 0 0 3L
0 36 3L 0 0 36 3L 0
0 3L L2 0 0 3L 4 L2 0
3L L 4 L2
2
0 0 3L 0 0
0 36 3L 0 0 36 3 L 0
36 0 0 3L 36 0 0 3L
3L 0 0 4L2 3L 0 0 L2
I 0 3L 4 L2 0 0 3 L L2 0
G = G1 = (62)
15 L 0 36 3L 0 0 36 3 L 0
36 0 0 3L 36 0 0 3L
3L 0 0 L2 3L 0 0 4 L2
0 3L L2 0 0 3 L 4 L2 0
EI L T d2N T d2N 2
T d N2
T
d 2N2
UE =
2 0 u dy 2 dy 2 u + w dy 2
1 1
dy 2
w dy
L T L T
EI T d 2 N1 d 2 N1 d 2 N2 d 2N 2
UE = u dy u + w
T
dy w (63)
2 0
dy 2 dy 2 0
dy 2 dy 2
25
1 T 1
UE = u K 1 u + w T K 2 w (64)
2 2
U E
= Kq e (65)
Para obter solues mais realistas, necessrio introduzir o efeito das tenses
de corte, que esto caracterizadas pela quantidade:
12EI
a= (66)
GSL2
onde:
E
G= (67)
2(1 + )
12 0 0 6L 12 0 0 6L
0 12 6L 0 0 12 6L 0
0 6L (4 + a) L2 0 0 6 L ( 2 a) L2 0
KE =
EI 6 L 0 0 ( 4 + a) L2 6L 0 0 (2 a) L2
3
(1 + a ) L 12 0 0 6L 12 0 0 6L
0 12 6L 0 0 12 6L 0
0 6L (2 a) L2 0 0 6 L ( 4 + a) L 2 0
6 L 0 0 ( 2 a ) L2 6L 0 0 (4 + a) L2
(68)
2.3 O MANCAL
Material
Viscoelstico
Mancal
Material
Viscoelstico
(a) (b)
Mancal
Massa do
Mancal
Material
Viscoelsti
Material
Viscoelstico
(c)
(d)
M
d m ( t ) N
d n (t)
( t ) + bm = E 0 (t ) + E n (69)
m =1 dt m n =1 dt n
M N
( t ) + bm D (t ) = E 0 (t ) + E n D n (t )
m
(70)
m =1 n =1
( t ) + b1 D [ (t )] = E 0 (t ) + E1 D [ (t )] (71)
( ) + b1 (i ) ( ) = E0 ( ) + E1 (i ) ( ) (72)
( ) E 0 + E1 (i )
E c ( )= = (73)
( ) 1 + b1 (i )
E 0 + E (i b)
E c ( )= (74)
1 + ( ib )
29
E c ( ) = E ( )(1 + i ( )) (75)
( ) G0 + G1 (i )
GC ( ) = = (76)
( ) 1 + b1 (i )
G0 + G (i b)
Gc ( )= (77)
1 + (i b)
30
ou na forma complexa:
Gc ( ) = G ()(1 + i ( )) (78)
Fu k xx 0 u
= (79)
Fw 0 k zz w
31
Sendo
k xx =LGc () (80)
k zz =3 LGc ( ) (82)
Devido a que neste trabalho foi considerado um mancal de rolame nto, a matriz de
rolamento, a matriz de amortecimento prpria do mancal ser desconsiderada. Por
outro lado, a rigidez elevada deste tipo de material faz com que seja considerado
apenas, pela condio de rigidez em srie, a rigidez da manta viscoelstica. Em
substituio, a matriz de rigidez do sistema, passar a ser complexa por serem
adicionados os termos de k xx e k zz complexos do material viscoelstico. Essa nova
matriz elementar agora dependente da freqncia, cuja responsabilidade direta a
manta de material viscoelstico nos mancais. A matriz elementar do mancal
viscoelstico adicionada no n correspondente do eixo.
33
LG( )[1 + i ( )] 0 0 0
0 3LG ( )[1 + i ( )] 0 0
Km = (83)
0 0 0 0
0 0 0 0
Disco i u i , wi , i , i
mancal 1 u1 ,w1
mancal n+1 u n + 1 , wn + 1
M1
M2
M3
M =
Mn
NX N
35
K1
K2
K3
K=
Kn
NX N
G1
G2
G= G3
Gn
NX N
36
Onde:
Tomando
f (t ) = {0}
(85)
q (t ) = {}e
st
s 2 M + s ( C + G ) + K { } = {0} (86)
ou
[D (s )] {} = {0}. (87)
38
Para que a equao (87) possua uma soluo diferente da trivial o determinante de
D(s) = 0.
2n 1
=> D (s ) = s
2n
+ p1 s + L + p2 n1 s + p2 n = 0
Para cada sj corresponder um {}j, de forma tal que [D(sj)]{}j= {0}. Portanto
q (t )
y (t ) = L (88)
q& (t )
2 nx1
C1 M M K M 0 f (t )
L M L y& (t ) + L M L y( t ) = L
(91)
M M 0 0 M M 0
y (t ) = e s t (93)
q( t )
y (t ) = L { } = L (94)
q& ( t ) s
40
[ sA + B] = {0} . (95)
B = A , (96)
onde:
= s = i e (97)
o autovalor;
s o coeficiente de Laplace;
B T = A T (98)
sendo:
- autovalores;
- autovetor a direita;
41
- autovetor a esquerda.
3.1.1 Ortogonalidade
i
Em Espndola (1990), os autovetores ortonormalizados so obtidos fazendo
ai
i
e , com a i = iT A i . Com estes vetores e as matrizes A e B, verifica-se a seguinte
ai
propriedade de ortogonalidade:
T A = [I ]
(99)
T B = [ j ] , (100)
= [ M s ]
T
. (101)
= [ M s ]
T
a) Primeira condio:
j k j T M k + j T K k = b j j k . (102)
b) Segunda condio:
( j + k ) M k + (C + G ) k = a j jk .
T T
j j (103)
j = j i j (104)
de ortogonalidade obtm-se
Tj (C + G ) k cj
2 j = = (105)
M k
T
j mj
e
43
Tj K k kj
j = = j + j = = = j.
* 2 2 2
(106)
Mk
j T
j mj
(f2).
44
0 0
L
G1 = M O M
0 Ip
L
0
Ip 0
q (t ) = e i t = e s t (109)
[ (M i G ) + i C + K ] = {0} .
2
1 (110)
q( t )
y (t ) = L (111)
q& ( t )
2nx1
q (t ) = e s t y (t ) = L e s t = e s t , (112)
s
46
Chamando M i G1 = M , tem-se:
[ ]
s C M M + [K M 0] = {0} (114)
e adicionando a igualdade:
[
s M ] [ ]
M 0 + 0 M M = {0 }, (115)
chega-se a expresso
C M M K M 0 0
s L M L + L M L = L (116)
M M 0 0 M M 0
ou, ainda,
C M M K M 0
i L M L + L M L = {0} (117)
M M 0 0 M M
ou, em forma compacta:
47
[ i A + B] =0
ou (118)
[s A + B] =0
onde
C M M
A = L M L
M M 0
K M 0
B = L M L .
0 M M
B = A (119)
sendo,
= s = i .
B T = AT (120)
48
M &q&(t ) + ( C + G ) q& ( t ) + K q( t ) = f (t ) ,
C + G M M K M 0 f (t)
L M L y& (t ) + L M L y (t ) = L
(121)
M M 0 0 M M
0
onde
q( t )
y (t ) = L .
q& ( t )
49
f (t )
A y& (t ) + B y (t ) = L = {f y (t )}.
(122)
0
y (t ) = p (t ) (123)
T A p& (t ) + T B p (t ) = T {f y (t )} (124)
ou
\
I p& (t ) + j
T
{ }
p (t ) = f y (t ) . (125)
\
p& k (t ) + k p k (t ) = N y (t ) ,
k
(126)
50
y
onde o valor de N k (t) para desbalanceamento encontrado atravs da seguinte
relao
0 0
M
fila 4 (i 1) + 1
M
Fu m u e 2 sin t
f y (t ) = Fw = m u e 2 cos t fila 4 (i 1) + 2
(127)
0 0
M M
0 0
[ ]
N k (t ) = mu d 2 T ( k ,4i 3) sen t + T ( k ,4i 2) cos t .
y
(128)
onde
C = mu e 2 , a1 = T (k ,4i 3) e b1 = T (k ,4i 2).
A + k B = b1C
(132)
k A B = a1C
b1C k B
A= (133)
k
a1C + b1C
B=
k 2
+
k
a1C + bC
B= 1
2
k +
2
52
a1 C + b1C k
B= 2
(135)
k +
2
(
a C +b C
b1C k
)
( )
1 1 k
2 2
k +
A= . (136)
Nota:
uma vez calculado os pk(t) com k =1,...,2n, calcula-se y (t ) = p(t ) .
q (t )
como y (t ) = L as primeiras n componentes representam a soluo das
q& (t )
coordenadas generalizadas q(t), a resposta do sistema a qualquer excitao do
tipo desbalanceamento.
3.1.4.1 rbita
f (t)
A y& (t ) + B y (t ) = L = L , (137)
0
q (t )
sendo y (t ) = L ,
q& (t )
y (t ) = p (t ) (138)
T A p& (t ) + T B p (t ) = T L (139)
\ \
aj p& (t ) + bj p (t ) = T f (t ) (140)
y
\
\
\
I p& (t ) + j p (t ) = T f (t )
y (141)
\
\ F ()
P ( ) = T f ( ) = T L
i I + j y
\ 0
ou
\ \
+ P () = T f () . (142)
i aj bj y
\
\
Conseqentemente,
1
\
P( ) = i a j + b j T f ( ) (143)
y
De forma compacta,
Y () = [ ( )] f y () , (145)
E = 2 x 1011 N/m2
= 7800 kg/m3 e
v = 0,3 (coeficiente de Poisson).
Os dois mancais so assumidos idnticos e caracterizados por:
Disco D1 D2 D3
Espessura 0,05 m 0,05 m 0,06 m
Raio interno 0,05 m 0,05 m 0,05 m
Raio exterior 0,12 m 0,2 m 0,2 m
Diagrama de Campbell
FREQNCIA VALOR DE
REFERNCIA
F1 55,408
F2 67,209
F3 157,90
F4 193,71
F5 249,90
F6 407,62
F7 446,62
F8 715,03
F9 622,65
F10 1093,00
FONTE: LALANNE E FERRARIS (1990)
58
3.1.6.2 Exemplo 2
Dados de entrada;
Esquema do sistema;
Diagrama de Campbell;
Resposta permanente no domnio do tempo a desbalanceamento;
Resposta permanente no domnio da freqncia a desbalanceamento;
Modos de vibrar.
A tabela abaixo mostra as dez primeiras freqncias naturais do rotor a 25000 rpm
apresentadas em Lalanne e Ferraris (1990) e as dez primeiras freqncias naturais
calculadas com o cdigo numrico em Matlab, desenvolvido no LAVIB, se observam
valores totalmente equivalentes, com erros menores que 1%.
F1 55,408 Hz 55,424 Hz
F2 67,209 Hz 67,243 Hz
F3 157,90 Hz 158,03 Hz
F4 193,71 Hz 193,95 Hz
F5 249,90 Hz 250,01 Hz
F6 407,62 Hz 408,33 Hz
F7 44662 Hz 447,83 Hz
F8 715,03 Hz 716,24 Hz
F9 622,65 Hz 624,29 Hz
F10 1093,0 Hz 1096,0 Hz
64
[ 2 M + i ( C + G( rpm )) + K ( )] Q( ) = F ( ) (147)
- existncia do termo i e
Q( )
Y () = L . (148)
iQ( )
[ ]
i C + G ( rpm ) M M Y ( ) + [K ( ) M 0] Y ( ) = {F ( )}, (149)
i [M M 0 ] Y ( ) + [0 M M ] Y ( ) = {0} (150)
C + G( rpm ) M M K ( ) M 0 F ( )
i L M L Y ( ) + L M L Y ( ) = L . (151)
M 0 0 M M 0
M
i AY ( ) + B Y ( ) = {Fy ( )} (152)
ou
66
[i A + B ]Y ( ) = {Fy ( )} (153)
onde
C + G ( rpm ) M M K ( ) M 0 F ( )
e {Fy ( )}= L .
A= L M L , B= L M L
M 0 2nx 2 n 0 M M 0
M 2 nx2 n
2nx1
B = A (154)
B = A , (155)
onde = i .
67
3.2.2 Ortogonalidade
j A k = a j
T
jk (156)
e
j B k = b j
T
jk (157)
= 1 se j = k
com .
= 0 se j k
jk
j = [ j M j j ] T
K ( ) M 0
k = [k M k k ] T e B= L M L
0 M M
obtm-se
j k Tj M k + Tj K k = b j j k . (158)
j = [ j M j ] T
C + G ( rpm ) M M
k = [k M k ] T e A= L M L
M M 0
obtm-se
( j + k ) Tj M k + Tj ( C + G( rpm )) k = a j j k (159)
j = j + i j . (160)
kj
j 2= = j = (i j ) 2
2
(161)
mj
69
e, portanto, j = i j .
2j = 2j = 2j (1 + i j ) (163)
logo
j = j 1 + i j . (164)
(
2j = Re 2
j )
(165)
( )
Im 2
j =
j
Re ( )2
j
Neste caso particular, em que o rotor est composto por mancais com material
viscoelstico, a matriz A que contm a matriz giroscpica funo da rotao do eixo
( rpm ) e a matriz B que contm a matriz de rigidez complexa e funo da freqncia
uma reta que cruza as curvas das freqncias naturais, sero extradas as freqncias
naturais do sistema de forma equivalente ao trabalho Espndola e Floody (1992) para
uma viga sanduche (metal - material viscoelstico - metal). Esse processo deve ser
repetido para todas as rotaes do rotor, resultando em um novo diagrama de
Campbell Final, agora ( j versus rpm), contendo as rotaes crticas do rotor
dinmico viscoelstico. A partir deste diagrama de Campbell Final possvel
determinar as caractersticas dinmicas do sistema rotor dinmico viscoelstico. Na
figura 19 apresentado um esquema mostrando a forma como o diagrama de
Campbell construdo. Na figura 19b representado o diagrama de Campbell interno
e na figura 19a o resultado final. Como pode ser observado, para calcular as
freqncias naturais do sistema para uma rotao constante, preciso resolver um
problema de autovalores funo da freqncia devido caracterstica da matriz de
rigidez.
71
j j
(b)
N rpm
1 = rpm = (166)
30
N
2 = rpm = rpm (167)
2 30
72
onde Nrpm a velocidade de rotao do eixo (rpm). Estas duas retas, utilizadas para
encontrar as rotaes crticas, representam s foras de excitao mais freqentes que
se apresentam na prtica: massa desbalanceada, cuja freqncia coincide com a
rotao do eixo (reta 1 ) e a excitao produzida pela instabilidade dos mancais
hidrodinmicos, a qual possui uma freqncia aproximadamente igual a metade de N
(reta 2 ).
Para uma excitao por massa desbalanceada, cuja freqncia coincide com a
freqncia de rotao do rotor ( rpm = ), o mdulo de cisalhamento
[ 2 M + K ( )] = F ( ) (168)
onde M = M iG 1 e = s2 .
Autovalores direita:
K ( ) ( ) = ( ) M ( ) (169)
[ 2 M + K ( )] Q ( ) = F ( ) (171)
onde M = M iG 1 .
Autovetores direita: K ( ) ( ) = ( ) M ( )
Autovetores esquerda : K T ( ) ( ) = () M T ( )
Q ( ) = ( )P( ) (172)
T ( ) [ 2 M + K ( )] ( ) P( ) = T ( ) F ( )
[ 2
]
I + () P( ) = T ( ) F ( ) (173)
Assim, no espao modal, as coordenadas generalizadas principais podem ser
determinadas por
P( ) = [ 2 I + ( )] 1 T ( ) F ( ) . (174)
75
Q( ) = ( )[ 2 I + ( )] 1 T ( ) F ( ) . (175)
e no domnio da freqncia,
[ 2 M + K ( )] Q ( ) = F ( ) . (177)
q (t ) = Q( ) e it + Q ( ) e it , (178)
onde
76
Q ( ) e i t o complexo conjugado de ( Q ( ) ).
3.2.6.1 rbita
i j k
u i (t ) 0 wi ( t ) = (u i (t + t ) wi (t )u i (t ) wi (t + t )) j (181)
u i (t + t ) 0 wi (t + t )
G0 = 3,57x106 Pa
G = 1,79x108 Pa
79
= 0,435
b1 = 2,46 x 10-3 (T0 = 273 K);
Disco
Mancal
Mancal
Eixo
Material
viscoelstico
Rotor
Dados do eixo:
Nmero de disco, ND = 2;
Raio externo, R = [0,280; 0,0445] m;
Raio interno, r = [0,025; 0,25] m;
Densidade do ao, = 7840 kg/m3.
81
Nmero de mancais, NC = 2;
Posio dos mancais, pc = [0,1835; 0,6435] m;
Massa dos mancais, Mc = [0,770; 0,770] kg;
Inrcia dos mancais (xx), Icxx = [0,00159018; 0,00159018] kg.m2
Inrcia dos mancais (zz), Iczz = [0,00690495; 0,00690495] kg.m2
Mdulo de cisalhamento e fator de perda dados pela equao (76), (79) e (80) e a
figura 22.
ROTOR
4 INSTABILIDADE
a1s 2
+ a2s + a3 = 0 (182)
s j = j i j (183)
seguinte forma:
jt i jt i jt
y (t ) = e [ e + e ] (184)
87
jt
y (t ) = e [cos t ] (185)
jt
e
2 cos t
VISCOELSTICO
11,8667 11,8830
19,4667 19,4828
37,0667 37,0824 40,2
47,5333 47,4821 51,7
59,6667 59,6152 71,8
112,4000 112,4139 77,13
159,0000 159,4793 192
243,1333 243,5439 371,8
281,8667 283,2096 376,73
416,4000 417,8062 588,66
501,8667 505,5374 594,86
662,8000 663,3334
6 RESULTADOS
Foi montada uma bancada experimental para validar a teoria apresentada, devido
mesma no apresentar condies ideais de medio foi realizada somente a
verificao visual das freqncias criticas. O rotor montado para a validao tem
capacidade para realizar medies somente at 60 Hz. Os resultados foram bem
prximos dos valores das rotaes crticas obtidas com o cdigo numrico prprio, o
que mostra um resultado promissor, mais ainda preciso realizar uma medio
precisa, atravs de sensores de medio para validao do cdigo numrico
desenvolvido.
94
7 CONCLUSO
REFERNCIAS BIBLIOGRFICAS
1. Bagley, R.L., Torvik, P.J., 1983, "A Theoretical Basis for the Application of
Fractional Calculus to Viscoelasticity", Journal of Rheology, Vol. 27 (3), pp.
201-210.
2. Bagley, R.L., Torvik, P.J., 1979, "A Generalized Derivative Model for
Elastomer Damper", The Shock and Vibration Bulletin, Vol. 49, pp. 135-143.
9. Ewins, D. J.. "Modal Testing Theory and Practice". Research Studies Press.
1984, England.
14. Lopes, E.M.O., Bavastri, C.A., Silva Neto, J.M., and Espndola, J.J., 2002,
Caracterizao Dinmica Integrada de Elastmeros por Derivadas
Generalizadas , III Congresso Nacional de Engenharia Mecnica CONEM, 10
1 13 de agosto de 2004, Brasil.
16. Marynowski, K., Kapitaniak, T., 2002, Kelvin-Voigt versus Brges internal
damping in modeling of axially moving viscoelstic web, International
Journal of Non-Linear Mechanies 37 (2002) 1147-1161, Poland.
17. Nashif, A.D., Jones, D.I.G., Henderson, J.P., 1985, "Vibration Damping ", John
Wiley & Sons.
18. Panda, K.C., Dutt, J. K., 1999, Design of optimum support parameters for
minimum rotor response and maximum stability limit, Department of
Mechanical Engineering, Indian Institute of Technology, Kharagpur 721 302,
ndia.
19. Pritz, T., Analysis of four-parameter fractional derivate model of real solid
materials. Journal of Sound and Vibration 195 p 103-115. 1996.
20. Shabaneh, N.H., Jean W.,1999, "Dynamic analysis of rotor- shaft systems with
viscoelastically supported bearings", Mechanism and machine theory,
University of Toronto.
22. Sowdon, J.C., Vibration and Schock in Damped Mechanical Systems, John
Wiley & Sons Inc., New York, 1968.
24. Wilson, C. and Sandler, J.P., 1991, Cinematics and Dynamica of Machiney,
Haper Collor Publisher.
98
Proceedings of the XI DINAME, 28th February-4th March, 2005 - Ouro Preto - MG - Brazil
Edited by D.A. Rade and V. Steffen Jr
Abstract. Nowadays the rotating machines produce or absorb large amounts of power in relatively small physical
packages. The fact those machines work with large density of energy and flows is associated to the high speeds of
rotation of the axis, implying in high inertia loads and deformations of the axis, vibrations and dynamic instabilities.
Viscoelastic materials are broadly used in vibration and noise control of dynamic rotors to increase the area of
stability, due to their high capacity of dissipation of vibratory energy. The widespread model, used to describe the real
dynamic behavior of this type of materials, is the fractional derivative model. By using the finite element method it is
possible to describe the viscoelastic dynamics of the rotor. In general, the stiffness matrix is composed by the stiffness
of the shaft and bearings. In this particular case, this matrix is complex and frequency dependent because of the
characteristics of the viscoelastic material that is part of the bearings. A clear and simple numerical code is proposed
to calculate the modal parameters of a simple rotor mounted on viscoelastic bearings. A methodology to build a
Campbell diagram (natural frequency versus rotation frequency) is presented. It must be built through an internal
Campbell diagram (natural frequency versus variable frequency), for which the stiffness matrix is a function of the
frequency.
1. Introduction
Nowadays the rotative machines produce or absorb larger and larger amounts of power in relatively small physical
packages. The fact those machines work with large density of flows of energy is associated to the high speeds of rotor
rotation. It implies in high inertia loads and potential problems with deformations of the axis or axle, high levels of
vibrations and dynamic instabilities.
Rotative machines often have problems of instability when working at high rotations, which can result in sudden
failures of the whole system or parts of it. This problem can be solved by including damping in the bearings. In general,
the area of stability can be enlarged with this type of control; also, the vibration levels can be reduced as well.
Viscoelastic materials are thoroughly used to control vibration and noise due to their high capacity to dissipate
vibratory energy (see Espndola et al., 2003). In order to do so, accurate knowledge of their dynamic properties is
essential to desing the control devices.
Several papers can be found in the literature, with the purpose of modeling simple rotors mounted on elastomeric
materials or on bearings made of this type of material. Generally, these papers use the model of Kelvin-Voigt, as
proposed by Shabaneh and Jean (1999), where the viscoelastic material is put under the bearings. This model has a
difficulty to represent the dynamic characteristics of most viscoelastic materials used in practice, when analyzed in a
100
wide frequency band (Pritz, 1996; Bagley and Torvik, 1983). It is noticed that this model is described by a differential
equation of integer order.
In the study accomplished by Marynowski and Kapitaniak (2002), the models of Kelvin-Voigt and Brgers are
compared in their capacity to describe the behavior of a viscoelastic material. The first is a model with two parameters
(spring and viscous shock absorber in parallel), and the second is described by four parameters. Similar results were
obtained for small values of internal shock absorption, but for the materials with larger coefficients of reduction the
model of Brgers proved itself more appropriate.
In Panda and Dutt (1999), polymeric materials are placed in the bearings. Using nonlinear optimization techniques,
it is possible to find the optimal dimensions to reduce the vibratory response of the system to unbalance excitations. In
Dutt and Toi (2002), models with three and four spring-shock absorber elements and integer order derivatives are used
to predict the behavior of a viscoelastic material that is part of a dynamic rotor. In that paper the aim was to study the
reduction of vibration and the changes in rotor dynamic behavior caused by the viscoelastic material.
In most of the papers mentioned above, the employed models of viscoelastic material cannot reproduce their
dynamic characteristics faithfully in a wide frequency band.
Here it will be presented a numerical model capable of predicting the dynamic response of a simple rotor in steady
state, whose bearings are made of layers of viscoelastic material. The model used for the polymeric material is the
fractional derivative model with four parameters, due to its capacity to describe the real dynamic behavior of the
material (Pritz, 1996). For this purpose, the characteristics of the viscoelastic material were determined by using the
methodology proposed in Espndola et al. (2003), Lopes et al. (2004) and Silva Neto (2004).
To describe the dynamic behavior of the system by Lagranges equations, it is used the finite element method. By
this way, the inertia (symmetrical and with constant coefficients), the gyroscopic (skew-symmetrical and a function of
the rotating speed), and the complex stiffness (composed by the stiffness of the axis and the bearings, which is
frequency and temperature dependent due to the viscoelastic material) matrices are obtained.
A simple strategy is proposed to calculate the modal parameters of the system. In this numeric implementation, a
Campbell diagram is generated, through which it is possible to determine the critical rotations of the dynamic
viscoelastic rotor. Due to the characteristics of the system - the complex stiffness matrix is a function of the frequency -
the final Campbell diagram should be obtained through another internal Campbell diagram. That is, when the rotation
of the axis is established, the inertia and the gyroscopic matrices are constant, but the stiffness matrix is a function of
the frequency for a given temperature. For each rotation, the natural frequencies of the system are functions of the
frequency, therefore, they should be found through another internal Campbell diagram (natural frequency versus
variable frequency). These steps follow the sequence presented in Espndola and Floody (1999), when the dynamic
behavior of a sandwich beam (steel viscoelastic material - steel) was studied.
To validate the numerical model, a simulation will be accomplished on a simple rotor system.
The simple rotor in study is basically composed by an axis, one or more disks and several flexible bearings, made
of viscoelastic material. The force upon the rotor can be caused by unbalances (synchronous excitation = rpm ),
instabilities of hydrodynamic bearings (asynchronous excitation 0,5 rpm ) or excitation through to the base by
rpm . This paper will consider only unbalance load.
The general equations of the rotor can be introduced through Lagranges equations (see Eq. (1)). So it is necessary
to define the kinetic energy T and the potential energy U of each element of the rotor, besides the virtual work done by
an external force upon the bearings.
d T T U
+ = Fqi (1)
dt q& i qi qi
In equation (1), qi is the i-th generalized coordinate and Fqi is the i-th generalized force. Then, using the finite
element method, it is possible to describe the rotor dynamics.
The disk is assumed to be rigid, being characterized only by the kinetic energy. Therefore, R0 (X, Y, Z) is defined as
the inertial coordinate system, coincident with R (xi , yi , zi ), a fixed system in the center of the disk (see Fig. 1). The
coordinates XYZ and xi yi zi are related by three angles, , and . To describe the rigid body rotation of the rotor
concerning the axis X, Y or Z, the Euler angles are considered (see Fig. 1).
The instantaneous angular speed vector of the disk can be written in the system of reference R as (Lalanne, 1990):
Considering linearity, the angles ? (rotation around axis X) and (rotation around axis Z) are considered small, so
that:
TD =
1
2
( 1
) (
M D u& 2 + w& 2 + I Dx x2 + I Dy 2y + I Dz 2z
2
) (4)
where MD is the mass of the disk, u and w are, respectively, the displacements in the X and Z-direction (see Fig. 1),
and, n this particular case, the disk is symmetrical, being verified that IDx = IDz , where IDx and IDz are the transverse
inertia in the X and Z directions. On the other hand, it is supposed that the angular speed remains constant, so & = =
constant. Therefore, the expression of the kinetic energy of the disk is:
TD =
1
2
(
M D u& 2 + w
1
2
) ( 1
) (
& 2 + I Dx & 2 + & 2 + I Dy 2 + 2 & + & 2 2
2
) (5)
TD
1
2
( 1
2
) ( 1
) (
M D u& 2 + w& 2 + I Dx & 2 + & 2 + I Dy 2 + 2 &
2
) (6)
In the previous expression, it is observed that the term (1 / 2 ) I D y 2 is constant, not having any influence in
Lagranges equations. The last term, IDy
& , represents the gyroscopic effect.
The shat is characterized by the potential and kinetic energy. The expression for the kinetic energy of the shat is the
result of an extension of the kinetic energy of the disk (see Eq. (4)). If the element has length L, its kinetic energy can
be expressed by the following equation:
(& )
S I
(u& )
L L L
TE = 2
+ w& 2 dy + 2
+ & 2 dy + IL 2 + 2 I & dy (7)
2 0
2 0 0
where I is the transverse inertia, is the density and S is the transverse area.
Considering the symmetry of the axis (Ix = Iz = I) and neglecting the effects of axial forces, the expression for
potential energy is defined by:
EI L 2 u 2 2 w 2
UE =
2
0
2 + 2 dy
y y
(8)
It is composed by two parts: the bearing itself and the viscoelastic material. The viscoelastic material can be added
between the external layer of the bearing roller and the bearing carcass or under its own base, as shown in Fig. 2a and
Fig. 2c. In the first situation, the inertia of the bearing can be neglected, while in the second, the inertia of the bearing
must be considered. Figures 2b and 2d show the simplified outlines for both situations mentioned above.
103
Viscoelastic
Material
Bearing
Viscoelastic
Material
(a) (b)
)
Bearing
Mass of the
bearing
Viscoelast
ic material
Viscoelastic
material
(c)
(d)
Figure 2. Model of bearings with viscoelastic material.
The viscoelastic material and the bearing are placed in sequency. Because of that, the stiffness of the bearing with
roller is higher than the stiffness of the viscoelastic material layer and the resulting equivalent stiffness will be that of
the viscoelastic material. The model used to describe the real dynamic behavior of the viscoelastic material is the four
parameter fractional derivative model. According to Bagley and Torvik (1983), the unidimensional constitutive
equation in fractional derivatives is:
M N
(t) + b m D (t ) = E 0 (t ) + E n D n ( t )
m
(9)
m =1 n =1
where b m , m , n , E0 and En , are the parameters of the considered material. The operators D m and D n represent
fractional derivatives. When M=N=1 and = , the equation above is denominated the four parameter fractional
derivative model.
(t ) + b1 D [ (t )] = E0 ( t) + E1 D [ (t )] (10)
The relation () / ( ) is denominated the elasticity modulus of the material (see Eq. (12)),
Ec ( )= ( ) / () = E0 + E1 (i ) / 1 + b1 (i ) (12)
E0 + E (ib)
Ec ( )= (13)
1+ ( ib)
104
where E1 = E b1 , b1 = b and Ec ( ) is the complex modulus of the material. In general, this modulus is a function
of frequency and temperature. In this paper, the temperature will be considered constant, so, it will not be included as
an independent variable. Still, the elasticity modulus Ec ( ) can be written in a general way by:
E c ( ) = E ( )(1 + i ( )) (14)
where E () is the real part of Ec ( ) , and () = Im(E c( ))/Re(E c( )) the loss factor. E 0 and E represent the
inferior and superior asymptotes of the dynamic modulus of elasticity. The exponent represents the slope of a
tangent straight line in frequency corresponding to the point of inflection of the curve E () , or, also, to the maximum
point of loss factor. The parameter b is the relaxation time.
The equation that represents the complex shear modulus is written in the following way:
Gc () = () / () = G0 + G1 (i ) / 1 + b1 (i) (15)
or, in the complex form, Gc () = G()(1 + i ()) , where G( ) = Re (Gc ( )) and ( ) = Im(Gc( ))/Re(Gc( )) .
For elastomers, ( ) = E ( ) = G ( ) (Snowdon, 1968). In the model considered herein, where the viscoelastic
material is added under the bearing (Fig. 2c), only the component of xx and yy of stiffness and damping will be
considered. In X direction, the stiffness will be represented by the shear modulus and in Z by the elasticity modulus.
Disregarding the stiffness associated to the rotations and ? (in Z and X direction, respectively) of the rolling
bearing, the stiffness matrix of the viscoelastic material will be given by:
Fu k xx 0 u
= (16)
Fw 0 k zz w
where L = A / h , A is the loaded area, h is the viscoelastic material thickness, and E a = E c . kT the apparent modulus of
elasticity. According to Nashif (1985), the apparent modulus of elasticity is obtained through the modulus of elasticity
and the form factor k T. In this paper, the layer of viscoelastic material was assembled in a such way that lateral
expansion is allowed, or k T 1, what implies E c = E a (see Fig. 3)
In the transition frequency of elastomers, it can be considered that Poissons coefficient is approximately equal to
0,5 (Silva Neto, 2004), so E a = 3Gc and the Eq. (17) takes the following form:
k zz= 3 L Gc ( ) (18)
105
According to the finite element method, it is considered that each node of each element of the rotor has four degrees
of freedom: two displacements u and w (in Z and X directions, respectively) and two rotations and around the axes
X and Z, respectively. Therefore, for node i, the generalized coordinate q i is represented by:
q i = [u i , wi , i , i ]
T
(19)
The application of Lagranges equations on the kinetic and potential energies of the elements of a simple rotor
(neglecting effects of circulation and stiffness caused by rotation), and assembling each elementar matrix conveniently,
results in the following differential equations (Lalanne, 1990):
( )
M q&&(t ) + C + G ( rpm ) q& (t ) + K ( ) q (t ) = f ( t ) (20)
where:
M is the inertia matrix (constant coeficient and symmetrical);
G is the gyroscopic matrix of the axis and disk (function of rotation and skew-symmetric);
C is the damping matrix of the axis and bearings (constant coeficient and symmetric);
K () is the stiffness matrix of the dynamic rotor and viscoelastic bearings (symmetric, complex and frequency-
temperature dependent). The stiffness of the viscoelastic layer is defined by:
LG ( )[1 +i ( )] 0 0 0
3 LG( )[1 +i ( )] 0 0
K ( ) m =
0
(21)
0 0 0 0
0 0 0 0
The system of equations that represents the movement of a dynamic rotor is given by Eq. (20). This equation is
valid for a sinusoidal excitation of frequency = rpm . To generalize this expression, it is possible to represent it in the
frequency domain, through Fourier transform. So,
2
( )
[ M + i C + G ( rpm ) + K ( )] Q ( ) = F ( ) (22)
where Q () and F () are the Fourier transforms of q (t) and f (t), respectively.
st i t
Considering the solution of the homogeneous ( f ( t ) = {0} ), given by q( t )={ } e ={ }e , the following
characteristic equation is obtained:
2
( )
s M + s C + G ( rpm ) + K ( ) = 0 , (23)
which corresponds to a polynomial of s 2n order. This polynomial has the following characteristics: there will be 2n
roots and those 2n roots will be complex and distinct.
106
To solve the eigenvalue problem, a transformation of generalized coordinates for the state space, Ewins (1984) and
Espndola (1990), is proposed. A new 2n x 1 vector of coordinates y(t) is defined.
q(t)
y(t) = L (24)
q& (t)
To represent the system of equations in the state space, consider the following equality:
resulting in
Ay& (t ) + By (t ) = {f y ( t )} (26)
where
C + G ( rpm ) M M K ( ) M 0 f ( t )
A=
L M L
, B= L M L
{
and f y (t ) = L } .
M M 0 2 nx 2 n 0 M M 2nx 2 n 2nx1
0
st
y (t ) = e
where = [ M s ] = [ M ] , and replacing it in Eq. (27), it follows that
T T
[sA + B] = 0 . (27)
Equation (27) represents an eigenvalue problem. Considering = s , this problem is defined by:
B = A (28)
where is denominated right eigenvector. Considering that A and/or B are not symmetrical matrices, the adjoint
problem of eigenvalues must be solved:
B T = AT , (29)
3.2. Orthogonality
[ j ]T A[ k ] = a j jk (30)
[ j ]T B[ k ] = b j jk (31)
107
= 1 se j = k
with .
= 0 se j k
jk
Starting from Eqs. (30) and (31), the following orthogonality properties in the configuration space are obtained (see
Espndola, 1990):
j k Tj M K + Tj K K = b j jk (32)
( )
( j + k ) Tj M K + Tj C + G ( rpm ) K = a j jk (33)
Equations (32) and (33) represent the orthogonality conditions as functions of matrices M, C, G and K. As the
values of are complex, they can be represented by the real part j and the imaginary part j .
j = j + i j (34)
The eigenvalues are complex and different; on the other hand, they are in some way related for being in the state
space. It is verified that the eigenvalues are formed by pairs j and -j . So, taking values of j k, but with j = ,
k
and applying these values in the orthogonality relations above, the result is:
kj
j 2= = j = (i j ) 2
2
(35)
mj
2
j (
= 2j = 2j 1 + i j ) (36)
4. Campbell diagram
In this particular case, where the rotor is mounted on bearings with viscoelastic material, matrix A , which has the
gyroscopic matrix, is a function of the rotation of the axis (O rpm), and matrix B is complex and a function of the
frequency (O). So, the eigenvalue problem is a function of the rotation and the frequency. That is, for a certain rotation
of the rotor (O rpm = cte), the eigenvalue problem is a function of the frequency and will be solved through an internal
Campbell diagram, j x , because K ( ) = K ( ) (1 + i ( )) . Starting from this internal Campbell diagram, considering
= j and using a straight line that crosses the curves of the natural frequencies, the natural frequencies of the system
are extracted in an equivalent way to Espndola and Floody (1999). This process should be repeated for all the rotor
rotations, resulting in a new final Campbell diagram, now ( j x rpm), containing the critical rotations of the
viscoelastic dynamic rotor. Starting from this final Campbell diagram, it is possible to determine the dynamic
characteristics of the viscoelastic rotor system. Figure 4 shows an outline of how the Campbell diagrams are built.
Figure 4b represents the internal Campbell diagram and Fig. 4a the final result. As can be observed, to calculate the
natural frequencies of the system for a constant rotation, it is necessary to solve an eigenvalue problem, function of the
frequency due to the characteristic stiffness matrix.
108
For the adjoint engenvalue problem it is necessary to consider that A and/or B are nonsymmetrical. The schematic
diagram below shows how the eigenvalue problem and the adjoint engenvalue problem should be solved for a rotating
system with viscoelastic bearings.
Frequency loop ( )
B = A
BT = AT
Internal Campbell diagram
In the final Campbell diagram, two straight lines can be represented by by the following equations:
1 = rpm = N /30 and 2 = rpm = ( / 2 )( N /30 ) , where N is the rotor spin velocity (rpm). These two straight
lines, used to find out the critical rotations, represent the most frequent excitation forces that happen in practice:
unbalanced mass, whose frequency matches the rotation of the axis (straight line 1 ) and the excitation produced by
the instability of the hydrodynamic bearings, which has a frequency approximately equal to the half of N (straight line
2 ).
A numeric example of a simple dynamic rotor mounted on roller bearings and layers of viscoelastic material is
presented. The dynamic characteristics of the viscoelastic material (pure butyl rubber) were determined in the
Laboratory of Vibrations and Acoustics of the Federal University of Santa Catarina (LVA-UFSC/PISA). A fractional
derivative model of four parameters was used to characterize dynamically the behavior of that pure butyl rubber. These
parameters are, for temperature T0 =273,0 K: G0 = 1,53 x 106 Pa; G = 1,11 x 108 Pa; = 0,396; and b 1 = 1,34 x 10-
2
.
Figure 5 shows the dynamic shear modulus and of the loss factor as functions of frequency and temperature.
109
Figure 6 shows the final Campbell diagram and the straight line 1 (unbalanced mass type excitation) that
determines, amongst all the natural frequencies of the system, the ones that participate in the response to the
unbalanced mass and are knwon as the critical rotations of the rotor. Given that the modal parameters of the system are
cut by this straight line, it is possible to find out the response of the rotationg system for this kind of excitation.
Therefore, it is observed that the proposed methology allows the determination of the dynamic characteristics of a rotor
the stffness matrix of which varies with frequency. This variation with frequency is due to the introduction of the
viscoelastic material under the bearings.
5. Conclusions
A simple and accurate methodology to find out the final Campbell diagram of a dynamic rotor, mounted on
bearings made of viscoelastic material, was presented and implemented in a numeric example. Due to the
characteristics of the stiffness matrix of this system, it is necessary to assembly two Campbell diagrams, one inside the
other, to determine the overall dynamic behavior.
The viscoelastic material used in the bearings was represented by the four parameter fractional derivative model.
A numeric simulation showed the steps to be followed to obtain the Campbell diagram. Once the dynamic
characteristics are known, it is possible to predict the response in distinct points and desing a control system, among
other control actions.
6. Acknowledgements
This work was partially supported by WEG, to which the authors are grateful.
110
7. References
Bagley, R.L. and Torvik, P.J., 1983, "A Theoretical Basis for the Application of Fractional Calculus to Viscoelasticity",
Journal of Rheology, Vol. 27 (3), pp. 201-210.
Dutt, J.K. and Toi, T., 2002, Rotor Vibration Reduction With Polymeric Sectors, Journal of Sound and Vibration,
262 (2003) 769- 793.
Espndola, J.J., Silva Neto, J.M. and Lopes, E. M. O., 2003, On the Measurement of Dynamics Properties of
Viscoelastic Materials, 2th Temtico de Dinmica e Controle da SBMAC, So Jos dos Campos, Brasil.
Espndola, J.J., 1990, "Projeto de Isoladores Aeronuticos", Final Report to the Bank of Brazil Fundation, Laboratory
of Vibrations and Acoustic, Federal University of Santa Catarina, Brazil.
Espndola, J.J. and Floody, S. E., 1999, On the Modeling of Metal - Elastomer Composite Structures: A Finite
Element Method Approach In: Applied Mechanics in the Americas - PACAM IV, v.8. p.1335 1342
Ewins, D.J., 1984, "Modal Testing Theory and Practice". Research Studies Press., England.
Lalanne, M. and Ferraris, G., 1990, Rotordynamics Prediction in Engineering, John Wiley & Sons Ldt. New York.
Lopes, E.M.O., Bavastri, C.A., Silva Neto, J.M., e Espndola, J.J., 2004, Caracterizao Dinmica Integrada de
Elastmeros por Derivadas Generalizadas, CONEM, Brasil.
Marynowski, K. and Kapitaniak, T., 2002, Kelvin-Voigt versus Brges Internal Damping in Modeling of Axially
Moving Viscoelstic Web, International Journal of Non-Linear Mechanies 37 (2002) 1147-1161, Poland.
Nashif, A.D., Jones, D.I.G. and Henderson, J.P., 1985, "Vibration Damping", John Wiley & Sons.
Panda, K.C. and Dutt, J. K., 1999, Design of Optimum Support Parameters for Minimum Rotor Response and
Maximum Stability Limit, Department of Mechanical Eng., Indian Inst. of Technology, Kharagpur 721 302, ndia.
Pritz, T., 1996, Analysis of Four-Parameter Fractional Derivate Model of Real Solid Materials. Journal of Sound and
Vibration 195 p 103-115.
Shabaneh, N.H. and Jean W., 1999, "Dynamic Analysis of Rotor- Shaft Systems with Viscoelastically Supported
Bearings", Mechanism and Machine Theory, University of Toronto.
Silva Neto, J.M., 2004, Identificao do Modelo Fracionrio de Materiais Viscoelsticos no Domnio da Freqncia,
Tese de Doutorado, UFSC, Brasil.
Snowdon, J.C., 1968, Vibration and Schock in Damped Mechanical Systems, John Wiley & Sons Inc., New York.
8. Responsibility notice
The authors are the only responsible for the printed material included in this paper.
111
C
CLLCCUULLOO DDOO FFAATTOORR DDEE FFOORRM
MAA
1 OB JE TIVO
2 ME TO DO LOG IA
I. E NS AIO DE C OMP R ES S O
O ensaio foi realizado para uma carga mxima de 2 Ton velocidade constante.
A duplicao da espessura de (A) para (B) implicou em uma reduo
proporcional na rigidez. Quando foram inseridas as ranhuras, como em (C), a rigidez
cai cerca de oito vezes. Desta anlise qualitativa, pode-se confirmar a eficincia de
almofadas com ranhuras no isolamento de vibraes.
8000
114
y2 = 1969x - 145,6
7000
y3 = 2027,5x - 73,816
6000
4000
3000
ENSAIO 2
2000
ENSAIO 3
1000 ENSAIO 4
1989,37 x 10 N/
G L + G H 0 [i T (T ) ]
G( , T ) = (1)
1 + 0 [i T (T ) ]
G ( 0,273) = 3,57 x 10 Pa
6
115
2) Geometria da Almofada
mm
- h altura da almofada
12 x 10 -3
Ea = 1989,37 x 10 3
2628 x 10 - 6
S S
k = Ea = E kT
h h
(3)
Ea
= kT
E
117
k T = 0,84
2
S
k T = 1 +
A0
(4)
2
2628
kT = 1 + 2
17726
Resulta:
k T = 1,04
V. SU M R IO DO S RE S UL TAD O S
Dif.
KEXPERIMENTAL Ea
GEOMETRIA k T EXPERIMENTAL k T TERICO % no
[N/m] [ N/m ]
kT
16311,50 x -
A Ea = 12,76 x 10 6 1,19 27,19
103
B 8811,15 x 103 Ea = 13,79 x 10 6 1,28 7,54 -
Como sugesto para a diminuio dos erros gerados na parte experimental, pode-
se realizar o mesmo experimento usando agora almofadas j vulcanizadas na
geometria (C).