April
1976
LATERAL VIBRATIONS OF .AN EXTENDmG ROD
by
Miodrag S. Jankovic
v.
16
.
•LATERAL VIBRATIONS OF AN EXTENDING ROD
by
Miodrag S. Jankovic
Submitted: . April,
1976
June,
1976
•
UTIAS Technical Note No. 202 eN ISSN 0082-5263
Acknowledgement
I express my sincere appreciation to Prof. P. C. Hughes for his help and guidance throughout the investigation and for the considerable time and effort which he expended on my behalf. Without his support and encourage-ment this work could not have been completed. I would also express my gradi tude to P .K. Nguyen for many helpful disçussions and comments.
This work was supported by the National Research Council of Canada under Operating Grant
8-7309.
TABLE OF CONTENTS
PAGE
List of Symbols iv
l·
INTRODUCTION 1..
II.
EXTEND TIJG PENDULUM 1..
2.1 Motion Equations 12.2 AA Approximation 3
•
2.3 The WKBJ Approximation 42.4 The Bessel Funçtion Solution when
t
=
0 72.5 Error Analysis(t
f
0) 0 82.6 Conc1usion 0 10
lIL EXTENDING CANTILEVERED ROD 10
3.1 Motion Equations 10
3.2 Dimensional Analysis 15
3·3 Contribution of the Second Mode 35
3.4 The WKBJ Approximation 35
3.5 Conc1usion 37
rY. References 38
•
..
•
B D E M N R e m n .. l.J t T p 9LIST OF SYMBOLS
Flexural stiffness Matrix defined by (3.1.12) Error defined by (2.5.1) or (3.5.1) Matrix defined by (3.1.18) Matrix defined by (3.1.19) Defined by (3.2.4) Extension rat~ Acce1eration of gravity Leng thMass of pendulum bob E1ements of M
E1ements of N Time
Nondimensiona1 time Mass per unit 1ength Lateral tip displacement Deflection ang1e
I. INTRODUCTION
The dynarnics of an extending rod has been investigated in Refs. 1 & 2. The equation of lateral motion of a rod whos e length changes wi th time has been derived there and was shown to be a fourth-order linear partial diffe-rential equation, under t~e assumption ofaxial rigidity and small deflections. The exact solution has not been found yet althcu gh a similarity solution has been found for specific axial speed history. ~uasi-modal* analysis was usedto transform the partial differential equation into an infinity of linear diffe-rential equations wi th variable coefficients. One solution was obtained in the form of Bessel function for the first mode only.
One purpose of this work is to assess the contribution of the second mode to the deflection of a flexible rod during extension, ani to find an approximate solution using tre first and second mode which will be more
appropriate then the Bessel function approximation. As a prelude, the analysis of an extending pendulum is done in section Ir because of the similari ty with the 'case of an extending rod and because of its simplici ty, namely only en e degree of freedom. This is equivalent in some sense to the one-mode represen-tation in the rod case because i ts length changes in a similar manner to that of an extending rod. This allows certain techniques and concepts to be evalua-ted prior to the more challenging case which has an infini te number of degrees of freedom. There is no exact solution available for the pendulum wi th arbi-trary extension rate, so certain approximations are exarnined.
The resulting error is defined in nondimensional form, and
plotted against nondimensional parameters to ensure generality of the analysis. The approximate analysis is tren applied in section 111 to the object of real interest - the extending rod.
II. EXTENDING PENDULUM
2.1 Motion Equations
Consider a simple pendulum of mass m and varying length
t(t),
as shown in Fig. 2.1~ Tt is constrained to oscillate in a vertical x-y plane aboutthe point O. Since the angular rriomentum about 0 is IDt28, and the moment due to gravity on m is -mgtsin8, the motion equation isd
dt -mgtsin8
Under the additional conditions that of (i) small angles, and (ii) the extention rate
t
is denoted bye, the governing equation becames*
The te rm I quasi-modal' was not us ed in the references; however i t is in troduced here as a useful nomenclature and eill be described presently.y
FIG. 2.1
'
.
..
8 +2
(e)' (g)
J
8.+J
8=
0Note tl1e pseudo - damping term. 2.2 An Approximation
The equation (2.1.1) does not a;ppear ta have a c10sed-form so:;Lution so we will attempt to find an approximate one in the form
8 = f(t)
cp(t)
(2.2.1)Substïtute (2.2.1) into (2.1.1) to get
.~
+ A(t)
~ +
B(t)
cp
=
0 (2.2.2)The coefficient A(t) and B(t) are defined as
(2.2.3)
The equation (2.2.2) has the same form as (2.1.1) but a subs·tantial silll]:llification can be made by choosing f( t) in s'Ucl1 a way: that
A=O
(2.2.4)·[Setting B
=
0 leads nowhere because one gets the same equatien for f as (2.1.1)].From (2.2.4), f(t) is faund to be
f
f(t) ::;
T
where f
=
const is re1ated to initial condition. oThe substitution of (2,2 •
.5)
into (2.2.2) gives·~
+
n2( t)
cp
=
0 where.
n2(t)
=
gi
~3
(2.2.5)(2 •
.
2.6)
Although
(22.6)
does not in general, have a alosed-form selection, the a.bsence of the cp term [c.f.(2 .. 1..1)]
is a considerable simplification.In particular , for constant extension rate e
n
2(t)
=='
.
]-g
+ et o
Even more particularly, if e
=
0 (i:"e., constant length) the distinction betweene
add cp is not necessary, and the solution for cp is elementarytcp
=
cpl..Cosnt+
cp sinnt2-where the consta..nts
cf>1
and cp depend on ini tial conditions • This does, however, suggest the idea tha;t(2.2.6~
might have an 'approximate solution of the form(2.2.7),
even when n = net) .One might suppose that such all approxi mation would be especially valid for small e. Denoting approximate solutions by a carat,("), and noting (2.2.1), (2.2,,5) and (2 .. 2.7), the approximate solution
/\.
e
=
chcosn(b) t+ C2sinn(t) t
:eet)
(2.2.8)will be examined. The constants C~ and C2 depend on ini tial condi tiom; in '
fac't,
j; e + e e
C~ == 1, o e
C
2 0 0°
0 n
0
..r
gij;"
e(t )"
eet ),where n 0
=
' 0,
e=
and e=
0 0
,
0 0The solution (2.2.8) is now shown to be qui te diffQrent in phase and atnplitude from the numerical solution which will be referred to as the 'exact' one. The difference grows larger as time t increases. Also the extension rate effects thequality ofthe approximation by making i t better for smaller values of e. 'Ihe numerical integration is performed using
tre
-R~ e-Kutta me th od (RKGIL) with s,tep-size :.H = 0.003. The calculation with H/2 gives the same result tof i ve digi ts ~d for 2H the accuracy is to three digit s D '
" Tt is obvious from Figso (2.1) and
(2.2)
that the approximat:ion, (2.2.8)
isnot ,a particularly good one even for 'e/(n 1, ) .. 060352~ One can", concludethat the approximation shoilld not be consid~r~d sati.sfactory for
el
(n 1, )>
0.0352, and that a search for a better approximation shruld beunder-ak
°
o - ,t e n . , ' ,
2.3 The WKBJApproximation
The WKBJ approximation mentioned by cunningham
3 ,
which gives an appr0ximate solution'for Eq.(2.2.6)
is now examined. It is assumed that((
....
FlG.U S 8 Ol I -DACT t-""-"O ... TlOIf,m. (2.2." nl I " \', ti! \ l ,-VUICT 1-N'f'IIII)ICI"""" m (!..UI ,8 i . c C.nt+C 'inOt S,.O.I roet &.-Q.I rad/a 1-0.2 ft/s j _ I.Oft !: ",,>0, " .0.005 .... FIG. Z2 6. c..codlt ... e.Û'Q, J L,OIrad 8.·-01 rodl'l • • Ot fth J_Lott 1-0.0'51: " .. 0.005 -Oi § 8 AG. 2.~ 8 8..
-a' FIG. 2.4 DAICT ..., nc ..., AIIfRI)X''''''''''' 9,8 i.CCOlht.~·t s.,QJ rad 8.--0.1 rad/, . . . 02 ft" 1.. LO ft &,."""" H" 0.003 1DCACT""'~""'~TIOtts.c .
.:.;
+C,aln; O"l g.-Q1rod 8.--OJ rad,. • -Ql ft/, t.· .. o ft &,.".030. " -0.00' -Oll.
-!
'" -e» F1G. 2.6EXACT AfC) TH[ . . . , 1I'P'IIIOXr1rM'1"IOH
8,ê
, ..
c.co.~:..;tlnt /100 .. "'" é.,,-0.1 roeI/, . . . 0.2 ft/I t..lO" '!: 0.0562 H .0.00', I
net)
>
°
a1ways • . The WKBJ approximtte solution is.. ,
where Cl. and
C~
are arbi trary c onst~s
andt
7fJ(t)
=
Jo
neT) d'r(2.3.1)
This approximation presupposes thatn(t) is slow1y time varying to the extent that
For constant extension rare e
f
0,t=~ + e t
o
7fJ
=
,g
.r
g (,J] -Jl )
e 0
ConseCluent1y, from (2.2.1), (2'.2.5) and (2.31) one gets
A
8
=
The first derivative of (2.3.5) may be expressed as
8
n
=
- [2sil1'ljJ
+
(e/nt) cos7fJ]CJ.+
[2cos7fJ - (e/n.e) sin7fJ] C22t
'
n
1/ 2where it has been observed that
n
= -
=-21
en 7fJ=n.
Using the ini tial condi tions
ê(o)
=
e
o ~.
8(0) ~e
o (2.3·.3) (2.3.4) (2.3.5) (2.3:6) ,..
..
the constants Cl, and C2 are easily derived from
(2.3.5)
and(2.3
'
.6):
.
C
2=
ion~/2
[
~: +~
(n:to) 8
0 ]
The WKBJ approximation is now shown to be an excellent one giving virtually the same results as the 'exact' ones obtained by numerical integration. The results are identical to five digits for the inertial condi tions specified
in Figs.
(2.4)
and(2.5).
The accuracy lessens for higher values of theexten-sion rate e, as expec ted. In view of tre condition
(2.3.3)
one would expectthe WKBJ approximation to be valid wherever
e
n/;
«
2.3 .•.•... (2.3.8)The Fig.
(2.6)
shows8
as a function of8/n
which is useful,being the functional relation of two nondimensional variab~es. For example,
at point M for which
.
8 8 ;::: 8M ' n-o :;::(~
)
,
o M ~ of
0any combinati?n of (
8
M,nO)
which gives(s/nO)M
is possible. Ifë~
correspondsto ~
0
then2 8
M corresponds to 4~
0
and so on. The relatione
=
8(
8/~0) is alsoused on the error analysis which will be carried out in Section
(2.5).
The errorwill.be defined as an integral at.a distanc~ between two points, M on the curve
(8, 8/n )
and M' _on the curve (ê ;~/n),
taken over time to h l l infini ty,multipl~ed by
Jn .
0o
2.4
~---~~~~~~~~~~---~---~~o The Bessel Function Solution when ~ - - - -:;:: 0. When ~ :;:: 0 the WKBJ approximation is of no use. [From (2.3.7),
the constants Cl. and
0
C2 are both zero, and from(2.3.5),8
is zero also.]The solution exists in the form of Bessel functions. Thus, for ~ :;:: et, (2.
ê
:
:6)
beco:rres
+ b cp:;:: 0 t
where. b :;:: gje and in Ref. ~ the solution is found as
7
where JJ. and YJ. are first-order Bessel fundions of the first and second kinds.
2.5
Error Analysis{.eo
r
0)
There is not a unique defini tion of the error con:ani t ted by the
process of approximately the exact solution. WE: shall us e the folloVll ng
defi-nition, which roughly corresponds to an integrated error in a 'nondimensionalized
phase space' whose coordi nates are (8,
êjn) :
2 1 \ 2
+
(8-8) ] dtt
-; .
Note .that the definition has been arranged so that E is already
nondimensional. One knows that E can only depend on the parameters of the
problem - namely.,
e,.e
,n. ,8 ,8 - and dimension analysis indicates thato 0 0 0
E
=E
.
(
rr
o 0
, 8 ,
o
~:
)
(2.5.2)
Tt is possible therefore to plot E as a function of each of these variables :
1) E
=
E(rr) ,
with ( 80,~
0 ) 0 0 0 fixed;.
2) E = E(8 ) ,
o with (~o
orr)
fixed; o 03)
E=E(~),
orr )
fixed.no
0 with ( 8 0 ,These are plotted in Figs.
(2.7), (2.8)
and(2.9).
They show approximate lylinear character. With respect to Fig.
(2.8),
the same plot was obtained byindependently varying e and.e as to be expected fram dimension analysis;
this seemed as a partial checf{ on the results .
The step-size used for numerical integration was H
=
0.01because it proved to be optimal among the tested ones. The trancation error
was shown to be negligible, results being the same to four digits for H = 0.01
and
H/2.
All programs were run in double precision .. Using(2.38)
and Fig.(2.8)
i t was obtained that the error E was ra ughly0.54
for e/nJ
==Jïtl"3;
E is bigger when e increases or.e decreases. 0
o
Figure
(2.7)
shows that E is proportional to 8 and it is equalto zero for 8
=
O. From Fig.(2.9)
one ge'ts bigger values flor E when 8increases. I~ is worth mentioning that the same plot is obtained by var~ng
8 and keeping
n
constant as by varyingn
and keeping 8 constant. Thus, the0.1 E 0.. 0.. 0.< 0.0 E 0.1 0.' 0."
0.'-..
CU QJ 0. ... 0.. 0.0 0.0.0 E 0.10 ,uo. 0..10 '.00 0..40 0..., 0.10 0..00 0.01 '0. 0.0t .
-0.0141-t:r:
o.on. ... Cl COl +(t) + c..In+Ct} 9. tI"" H -o.Ot J...o.lrad ~.QO"'5 § • Clco.+(~~ ~nt(t) H. QOI...
8.-QI roef ~.o.l4S §. C1Cott(tÖ:'int(t} H _0.01 0.. FICl Z.9 AG. 2.7 lO AG. 2.8 LO...
-I. Te 0 •• • lt;L t.Oexpected results are proved quantitatively and qualitatively. 2.6 Conclusians
The approximation (2.2.8) is not acceptable even for e/nt
=
0.0352.Nevertheless, it has a very simple form and its accuracy increases for smaller
values of e/nt until i t becomes equal to the exact solution for e/nt = O. The
WKBJ approximation is an excellent one for e/nt
<
.Jï6/3 and i t is much superiorto (2.28) for all values of e/nt
r
o.
However, i t requires the integral (2.3.2)to be found in closed form (which is possible in our case). Otherwise we have
still to 'U3 e some numerical methods which may be as laborions as solving the
basic equation (2.1.1) numerically. The Bessel function solution should be
used for
t
= O.o
III. EXTENDING CANTILEVERED ROD
The object of this study is to treat an extending rod. The treatment of a simple pendulum made in the last section il1ustrated certain concepts ascertained the accuracy of candidate approximations.
3.1 Motion Equations
It is shown in Ref. 2 that the motion equation for a uniform
rod of mass/length p and flexural stiffness B, extending at uniform rate e is
By"" + p(y' + 2e
y'
+ e2 y")=
0 (3.1.1)Primes denote differentiation wi th respect to the axial coordinate x; overdots are time differentiations, and y is the small lateral deflection (Fig. 3.1). For a cantilevered rod, the appropriate poundary condtions are:
y(
0,t)
o
y'(O,t)By"(t,t) = 0 By"' (0,
t)
(3.1.2)One approach to the solution of (3.1.1) will be called the
'quasi-modal' approach. In this approach, it is recalled that when e
=
0, the solutionof (3.1.1) takes the form of a series of 'modes' in which each mode shape ~.(x)
oscillates sinusoidally in time for a nonextending rod 1.
00
y(x,
t)
=I
i=l
~.(x,t) f.(t)
1. 1. (3.1.3)
The dependence on the parameter
t
has been explicitly indicated in anticipationof the next step, which is to apply (3.1.3) even when th.e length
t
istime-varying - the quasi-modal approach.
The functions ~.(x,t) in (3.1.3) are defined by r
y
CP.
(x,};)=
11ïi [CCOSZ.X - coshZ.x) - d. (sinZ.x-sinhZ.x)n
1. "JIJ 1. 1. 1. 1. 1.
J
(3.1.4)
where
Z1.' = E. /}; and E. are the roots of the transcendental equation
1. 1.
COShE COSE + 1
=
0 (3.1.5)Thus El.
=
1.875 •.• , E2=
4.694, ..• and so on. byThe
a.
in (3.1.4) are defined 1.a.
1.=
COSE. + coshE. 1. =I-sinE. + sinh€. 1. 1.The orthonorma1ity properties of
cp.
(x,};) are noted: 1.These functions satisfy the ordinary differentia1 equations
cp
~'"1.
z'?'
1.cp.
1.=
0 with the boundary condi tions" CP.(O,~)
=
cp~(e,~) = cp~'(};,};) = cp~,,(};,};) = 0 1. 1. 1. 1. (3.1.6 ) (3.1.7) (3.1.8) (3.1.9)The quasi-modal approach to (3.1.1) is to substitute (3.1.3) even t}'lQugh
}; =
};(t). It is not c1ear that tnis procedure can be rigörous1y justified, a1-though one would expect useful resu1ts for same range of values of e>
O. When one does so, one obtains (Ref. 2):when the motion f contains the functions f.: 1.
,I
(3.1.10)
and the matrices _D, Mand N have e1ements. (d;J'}' (m .. } and (n .. } defined as .Jo ~J ~J
t
m ..=J
[-cp./2
+(t-x)cp~]
[-cp./2
+(t-x)cp'.]
dx ~J 0 ~ ~ J J =fot
n .. ~Jcp.
~[-CP
J./2
+(t-x)cp '. ]
J dx -.Moreover , D is diagona1, ~ is symmetrie, and N is skew-synnnetric;
d ..
=
0 ~J m .. = m .. ~J J~ n .. = -n .. ~J J~The numerical va1ues of these matrices are
M = N = 0.504995 0.527068 -0.559038 0.000000 0.654848 0,228501 0.527067... -0.559042 •.... 4.087497... 3.441512 ...•• 3.441515... 13.93843 .••••• -0.654852 ..••. -0.228555 .•.•• -0.000002 ....• -1.636973 .•... 1.636984 •...• 0.000028 .•... (3.1.12) (3.1.13 ) (3.1.15) (3.1.16) (3.1.17) (3.1.18) (3.1.19)
In view of (3.1.15) - (3.1,:"17) it is evident that the numerical va1ues of Mand
~ in (3.1.18) and (3.1.19) are accurate to at least four significant digits.
Thus it remains to solve (3.1.10), a set of 1inear ordindary
differentia1 equations with variab1e coefficients (sinee
t
=
t(t) ). They are inpractice truncated to a finite number of modes. I t remains to get the initia1 conditions f. and f. at t
=
°
from the :lnitia1 conditions y(x,O) and y(x,O). To this end/- it isJ.noted from (3.1.3) ·that.00
y(x,o)
= \ 'L
cp.(x,ft )f.(o)
J. 0 J.
i=i
; 1
Multip1ying through by
cp.
(x,ft o) and integrating yie1dsJ . ft f.(O)
=
J
y(x,O) CP.(x,ft ) dx J. J 0 o ' Simi1ar1y, we ca1cula:te 00Y
(x,t)=
L
[cp. (x,ft J. OJ. )f. + cp. (x,t) J. f. ] J. i=l where (x,t)It may be shown from the form of (3.1.4) that
(3.1.20)
(3.1.21)
(3.1.22)
(3.1.23)
Next, multip1y (3.1.22) through by cp.(x, ft) and integrate over x from
°
t/o ft • JUsing the orthonormality conditions, one obtains
a:t
t = 0,.
.
J
to
f
.(0)
=y(x,y)
cp.(x,
ft ) dxJ 0 J 0
(3.1.24 )
wh~re the f.(O) are a1ready avai1ab1e from (3.1.21). · This comp1etes the
'
.
The series soluticn, (3.103), must be truncated to a finite' number
of terms for practical reasons • . The severest possib1e truncation is to çn e
I me de I , . for which the differentia1 ~quation for fl. is a particular case of
(3.1.10), name1y
•
(3.1.25)The function y
=
Y = CPJ.fJ. is p10tted* fOT specified initial conditions andparameters in Fig ~
(3.1).
3.2 Dimensional Analysis
Equation (3.1.10),can be stuOied mere efficient1y with the aid
of nondimensional,ized variables .Lengths wf11 be nondimensiona1i~ed by
t
(assumed
f
0), anel times by the t;i.me factor. 0(3.2.1)
Thus a nondimens;i.o:pa1 time, T, and a nondimensional extension rate, e*, are
defined by T
=
tlt* e* = et*/t o (3.2.2) (3.2.3)The ratio of the instanta.neous 1ength to the initia1 1ength
t
is defined as R:o
R
=
j-
=
1 + e* (T-To) o
With these definitions, Eq. (3.1.10) becomes
where (.) now denotes (iifferentiai;;ion wi th respe ct to T, and
f*
=
!Jt
o3/
2(3.2.4)
(3.2.6)
* Tt appears that the authors of Refs. 1 & 2 have inadvertent1y made their graph
scale ten times the true va1ue for t ~ 0.5.
YI
0.1 Ist. MODF.p
= 108 f~ , 10 = O. ft to=
0.5 secti
(tol = I. ft 3/2 fdfol = O. ft 3/2 .J
= 1.56x 108JS
H.'" 10-51.8C Hp= 50H•
r
.~-....c::::::+---
~.-. O • '4lII\i!\I'M/4~. "'\NVV '''\I\ {\,(\ ~ ~C7 A . C).'~ 7 " '>. ...~ ' ==--=-=-f~,
_._.,
.
:::::::===-'""--=r==
,
o
0.1 (>:3- 0.4 0.5 0.6-'- 0.7 _.-0.8 Q9 LOt
.
·-0.1 FIG, 3.20I~
r
q N ~ ~.!::;
~ ~-r
t~
LIJ C) ~...
x LIJ Z LIJ :J:...
Z 0 Z~
Z ;:: +- Z 0 U ~ ,..;ci
LA:CD N ~ N ~ C\! IC) N +-u N rt) <.!) ~
..
As to the nondimensional form of
cp.
,
define1. cp~ = .,{}; o CPi 1. ~ = xl}; 0 (3.2.8) As mentioned earlier,
~.
--
(
e* ) (cp~ + 2 ~cp
~ * )2R
. 1. 1. 1. (3.2.10)where (') denotes differ·entiation with respect to ~.
The one-mode approximation to the solution, given dimensionally by (3.2.25) is
(3.2.11)
while thetwo-mode approximation written ou~ is scalar form is
(3.2.12)
(3.2.13)
Recall the elements m .. and n .. were given in Eqs. (3.1.18) and (3.1.19).
l.J l.J
A numerical study of (3.2.11) - (3.2.13) is now undertaken to assess the contribution of the second mode to the solutian. As a measure of the size of the solution, we use the tip deflection, y($, t). As a step in
this direction, one can ask; if either of the first two modes is excited what is the suppliëd exci tation at. the other via the coupling terms in
(3.2.12) and (3.2.13)?
To answer that question, the initial conditions of each of them are confirmed to zero, in turn. If the second mode is not excited the following
two cases exis"t. * (0.5) * (0.5) fl. :::: 1. fl.
=
O.
* (0.5) * (0.5) f 2=
O.
f 2=
O.
'* fl. (0.5)=
o.
(3.2.14) fi (0.5) .*-,=
1- (3.2.15) .* .*, f 2 (0.5)=
o.
f2
(0.5)=
o.
*
.
where fl. (0.5)
=
1 in (3.2.14) and f*(0.5)=
1. in (3.2.15) are set up for convenience. The ini"tial value for.T=
0.5 is chosen because smaller values require much smaller s"tep-size and coRsequentlY "the compu"tation cost increases. If T=
0 instability in integration occurs in some cases. Since'thedifferen-tial 0 equations are linear, any case in which the fir st mode is exci ted may be writrter). as a superposition at (3.2.14) and (3.2.15). Similarly, the infl.'uence of the second mode on the first one was assessed by using "tb:! following initial condi tions • * * (0.5) fl. (0.5)
=
O.
fl.=
O.
* * (0,5) f 2: (0.5)=
1- f 2=
O.
.* (3.2.16) .* (3.2:16) fl (0.5) ;::::O.
fl. (0.5)=
o.
'* (0.5) ;* (0.5) f 2=
O.
f 2=
1.
.
*
.
Tip deflection due to first mode,
yt,
and due to second mode, Y2 are plotted for the interval 0.5<
T<
20. The upper bound in,this ,ihterval is proved to belarge enough for the purposes of "the anlaysis'.
extension rate:
The Eqs. (3.2.12) and (3.2.13) are now examined for the following
1) e* = 10-4, which is the practically interesting case;'
2) e*
=
1, which gives a very'highvalue of exteneion rate for ordinary values of ~/p iof the order of 10 ft/s for ~ 0 = 1 and B/ P. .... 10 ff;t~I
rP
~).
,
3) e*
=
100" which is of academie in"terest only.For the
ers
satelli te, the \vfue ofBI
p is~bout
106 f.t/ s',' S,o, thata nondimensional extension rate of e*
=
10-, one gets from (3.2.3) e=
0.1/~ .ff,/sec. In order to obtain smoo"ther graphs for 0.5 ::: T
<
20 the step-~ize 0H
= 0.003 was chosen. For the initial condi"tions (3.2.l4) and e*
= 10-
4 the"tip deflections shown in Fig. (3.3) - ~3.61 are ohtained. The deflection l1as oscillatory character with frequency Ol. (Ol. will be defined precisely later) and almost constant amplitude as expected. The influence of the fir~t mode on the second one, which is observed through y~, is of the order of 10-0 and has
oscillatory character also. If there were no coupling terms, y~ wruld be zero for all';"time.
•
y; { ., ::::~.!
f~Ct·.) -I. ;:(1'.) -0. 1 - I~ 10 .. 1' I.!
,; 11'.1 .0. 2 model ti (T.'. 0. f; (T.o,-a.O:f
t
\,
r
\,
1
~
\
) ,\ I \
~
\, ( \ ' ,,\
~
,
)\
{ \ l
\
k
'
,
~.
J --0.& AG. l!.3 f~ (Te) • I I;(T'/';I
r. (1<.) • o. ? model ij ('re) • o.'I:
{.1. . .. 10- 4 \ Te .o.~ I mode f~(T.). I. I,"(T.). O.(
t
11 " 'li
l
11 11 11 11 • ..;..-l-+++ I I I I I I I I I I I I I I I I I I ~d~ I ~~ I I I I d~ I I I I 1 y; FIG. 3.4 Y. . .0-. {. ,.*.
KT4 T.t-0.5 fl~(T.). I. f.a(T.). o. 11* (Te)-Q i.' «.).0..
all, \,I \ 1
f
f \,
r \\
,Ilr!
\
1 \
~
\\' I, \
'
I, \
~
'T ' , tD·-FlG. 3.5 FIG. 3.F {.,
e*.10-4 1'.-0.5'''(1'.)
-
I. '1"(1'.'-0. I,·(T.)'O ;t(T.).O .'
~
.
Y.* . * The phase diagram
(ii,
y~),
Fig.(3.4)
is en ellipse but(Y2, y~) has a rather strange shape altho~h it is bounded, Fig.
(3.6).
FromFigs.
(3.3)
and(3.5)
it is obvious that Y2 is negligible in comparison to yfwhich proves very small coupling. If those coupling terms are neglected the
following equations are obtained:
"* *2 *
fJ. + n~ fJ.
=
0(3.2.17)
when the rotation
(:1 y
(~*
) 2 = . mJ.J.(3.2.18)
(:2y
-c;Y
m22 is used.*
*
The graphs nJ. 2 and n22 are plotted for different values at e*
in Fig.
(3.7)
and (~.8). It is obvious from these figures that there are somevalues of nf2 and n22 which are negative so it shoul~ be clear that the notation
(3.2.18)
does ~ot imply positi ve values of n12 and n22 for ' all values' of T. ; t •Both nf2 end n22 tend to zero when T --700 which shows asymptotically
non-oscillatory behaviour. The frequency n; is considerably higher than ni, as expec~
'ted.
The equations
(3.2.12)
and(3.2.13)
are rewritten to help clarifyFigs.
(3.6)
ani(3.10):
e* .* W*2 * - 2n1.2 R f 2 ·- ,2 f 2(3.2.19)
e* .* w~2 * - 2n2J. R f~,.
fJ. where(~*)
2 ( Illl.2 + nJ.2) =Figure
(3.6)
may then be explained as the phase diagram of+
o
n" I IQ -IQ
..
n. '-·0 - - - -__ .~..!'ïO=4'-" ---_._----FIG. 3.7 e -=10-2..
.
T FIG. 3.81 CI ... do
~~~~1~~~~
." d.Jod _~~ ii~î .:.~':r~~!:-!:I
0'" d d::~~~~~~
g " i.*
Y2
o
y, ( 0' "'10-' T.: o.~
f:
('r.). O. 2 rnodH fi(,..l-o. i~(To) -.. i;(To) -0. f~ (Tol oal , _ ;~ (t"J .. I. Jf\
A
i\ (\
0.1 1 LI
'
I 'I
"
1 r. I, ,f \,
I
'
J:
10.,.I
Ii
I/1,
1',r \,
I '1
,\
I,I,
20., ,.
-0.1 I . I 1 , , I . \ 0.1 T FIG. 314 f~ (T,!o 0 I Ir tr.) • ,:, I modiy;
l'1G. 3.'5 \V
\\;
I T." 0.5~.:
'10-'I
1;(To)-o. 2 _ 1;(,..)-0. ;,'(To)_" +;tTo)'O. y;y;
'0" ( 0' '-"10-4 Te" 0.5 fl·{t'.)·Q f,*(T.)-o. t,* (T.).'. f.*(T.,) .0. T-
I
i
~~~~,n, '~f~yl ~~~~
I
I
'
P
W
'1'
I
11
~
I·
I ,
'
" "
.'
-16&' FIG. 3.16y;
l~
'"
( 0' '*.0- 4 1'." 0.5 ft (To)-0. h*(T.). O.',0
(To)O .. I: (To)-o· ocr· y.y; (., .... 10- 4 T.= 0.5 ,,*(1ö.): 0. fa*(r.},",O-t.-(T.),",o. f.·CT.). I. FIG. 3.18 y; ( .
,
e-= 10-4 T." 0.5 f~ (T.)" O. 1"(,-,)=0. i,·(T.)' O. ft*{r.)"'I. T y , {.,
Y; .~::~~. ~I· (T. ).0f,* (T.l·o. .
!
2 modes',*
(T.).o. ; ... (T.)-I. ,; (~l '"0
'
1'
mode f; (T.) .. I. Ol T "IJjIil.DI\I\f1i1j' ~~~\!IMf\AAM~WiIiI\MI\IIAAMI\I\Iit;:-+----+-~ _o~[1m!1n t T , ; (TJ • 0.I,
mode fi (Tel -I. FIG. 3.20 y;~:'IO-'
!
T.,.. 0.5 ,; (1ö.) .. D. 2 modes ft*(T.l-O. fl-(T.,) .. o. i"(T.l. I. Y. FIG. 3.2'.
" ;:!H
... "" • ..;!.: /? , , "-., . "':; i a ol ...._,
ol 11 ol ~.
..
.
... d cl Cf~~!.~~l~
•
"
... cl cl ó~i~~~~~
.
.
,
''':;
/
I
...
_.J(I.e.t:. .:.,: ï'.:.:..-\
d~dd~~
·
~~~~~
.
4
',:: '':: " ..;''':; '':: .. ,::
I
;;~ll w· ..... ..:!.= '",..
..
d ó·~ ~I
~ I • i:,;. '';' \ \
\
'", Cf d d ... ;-:~ill:; ..., .. .,\:-vs.:~ ~~ _"~;H ':',;'.::~ Cf IS Cf ..: ~,~~~~~ '~ '':: a • y 8,..
""doo _s~~i]i .:Ä.i.-=~~-'o!' '':
..
_hH
.:,.;;!.: ... 9 ',;: • 1•
i•
I
Ir
l
;
t
9.
..
..
; ~..
..
'':: , / ol\
\
\ J r" ! • !.
"~~~
.
~É
",;i; .. ~ ... d '1 \ \ • 1/
/ /y~ ( ., '-.10·" T.ao", TI' tt.,· U. t,* (~). 0.
f.-
(T.). Q 0It f.-''tol· 0. '.D (T.)-I. f,-,1'<o) -I.',*
tr.,: I. i.-lt.) _0.·
AAA
Ar.r'il
·
AAr
,
~
I
Qt , ,. 0 0 : .'·
V
V
V
vV
J
\I
V
V
V
V
....
·M AG. 3.~ j: FG, 3. 57 { ., ".10,4 'f.ao.5 ,,·(r.).o. ;~:~:~:~: ft-tT.)· I. y; "'-T Y·.'f.+Y; :..
.
(. , '-·10T._a.5 -" ''-('1".)-0. 't9 (T.).a. 1:( • .).,. i.-h.).'. j; Q' -0.& AG. 3.~' ( ., ' •• 0-4 .... O~ f,4(T.). O. ,,·(,..).0. fl·lT.)-'. ,,*(T.)·I. y:I
~1\!\~~'1rlili
:
.
I
\
I ( '\
· ·
\ I
\
[I
1
\'
·
;
0., -OlI
/
r\·
Jp,~~,~
FIG. 3.51 y; ( ., ft. 10"4 "·0.5 ""{1".).o. ''-('1'.'-0. I,-tr.)-.. OR I,Dh'>_" ft (t;J -0. f,-(1OJ .0. ""h"I.,o. irDlr.) a •• flG. 5. 56 { ., ".10"4 T._O", '''(T.)Oo. ';(To)·o. ''-(T.)O', t.*tr.) a I. y' T•
.
*
*
w:Lth the frequency
n
2 , to which is added the existing frequency w1 , (Fig. 3.11).A similar explanation applies to Fig. (3.10).
For e*
=
1, the coupling terms in (3.2.12) and (3.2.13) have muchmore influence so that the approximation (3.2.17) is not accurate for all T.
Figures (3.21) and (3.22) show a similari ty at the beginning of the inte rval as
db Figs. (3.28) and (3.29). Referring to Figs. (3.7) ~nd (3.8) one can see that
n;2
andn~2
tend to zero much faster than for e*=
10-4. So, asymptotically,yf and y~ became straight lines much earlier as i·t is shown on Figs. (3.21)
-(3.39), and the oscillatory character of yf and y~ disappears. An important
conclusion is that the nonexcited mode is not negligible anymore being of the
same magnitude as the exëited one. Compare Figs. (2.21) with (2.23), (2.26)
with (2.28), (2.31) with (2.34) and (2.36) with (2.38).
The results for e*
=
100 are not of practical interest. For B/P ... l06~~jI:!S2: the extension rate is of the order of 10
5
/t
ft/S which is very largeand far from extension rates encountered in practice. 0 The approximation
(3.2.17) are not valid anymore because of the considerable coupling influence. [Cornpare Figs. (2.41) with (2.42), (2.44) with (2.45), (2.45) with (2.46),
(2.47) with (2.48) and (2.49) with (2.50)].
3.3 Contribution of the Second Mode
The contribution of the second mode is shown to be negligible
for e* =:
~06~
.:which gi yes the real extension rate of the order 0.1/t
0 ft/secfor B/p ~O. f~/s.The coupling effect is very small so that the same graph
is obtained for ini tial conditions (3.2.15) and (3.3 •. 1)
* (0.5) fl.
o.
* (0.5) f2 = O. * (3.3.1) fl. (0.5) =: 1. * (0.5) f 2 =: 1.*
Also,the contribution of the second mode Y2 , Fig. (3.55), is shown to be the
same for initial conditions (3.2.15) étnd (3.3.1). The amplitude at Yl. is about
five-times biggerthan the amplitude at y~. The contributim of the second iHode
.increases wi th time. However , for e* =: 1 and e* =: 100, the contribution of
the second mode cannot be neglected.
3.4 The WKBJ Approximation
The WKBJ approximation can be used for Eqs. (3.2 .• 17) and
conse-querrtly the first and second mode approximation are gi ven in explicit form:
*
y.
1. (3.4.1)
where i = 1 for the first mode, and i = 2 for the second mode. ~i is given by
=
JË.
1 e*n.
was def.ined by (3.2.17) and1
*
Q. dt=
1:it
(:~ )~c
cos (
1 1B.
1 E. 1 arc cos(B.
2.R
E. 1 B. 1 E. 1The constants Cl.i and C2i are found from the initial conditions
Y*(T ) o
=
Y
0 = (3.4.2) (3.4.2) (3.4.3)y. was calculated over the interval 0.5
~
T~
20, for e* =,10-4
and gives an1
exc~llen t accuracy having the same four significant digi ts as the I exac t'
solutions. From Figs. (3.7) and (3 .. 8) i t was concluded that the WKBJ approximation should be a good one for all·.T
>
1 when e*>
10-2 and for 0<
T<.co
whene*
<
10-2 i.e., for all T for whichnf
is slowly varying. ~-The error ana1ysis in paragraph (.2.5) can be applie d to the a.pproximate solution (3.4.1). The error is defined as
n .
--- - - - --- - - --- - - - --- - -- -- --- ---- --- - - - . . ,
-and similar lin~ar functions of (3.5.1) can be obtained for different variables
ejn ~ , ~ and ~jn
.
0 0 0 0 0 3.5 Conclusion
It ts evident that for extension rates characteristic of actual
rod-like (e*
<
10- ) the coupling between the modal coordinates can be neglectedas an excellent approximation. Therefore, if the ini tial conditions on the rod
indicate that only the first mode is initially present [consult Eqs. (3.1."21)
and (3.1.24)], then a one-mode analysis is adequate. Similarly, even if both
modes are present the coupling between them can be neglected for e*
<
10-4,
although the numerical integration of the coupled equations, (3.1.10), is not difficult.
Whenever, a single-mode equation is used, the WKBJ approximation to its seleçtion is an excellent one if an approximation is desired.
• 1. 2. Leeeh, C. M. Tabarrok, B. Leech, C. M. Kim, Y. I . REFERENCES
"Dynamics of Beams Under the Inf1uence of Convecting Inertia1 Forces", Ph.D. thesis, University of Toronto, Department of Mechanica1 Engineering, 1970.
"On the Dynamics of an Axia11y Moving Beam", Pergamon Press, 1974.
3·
Curmingham, W.J. "Introduction to Non1inear Ana1ysis", McGraw-Hi11, NewYork, 1958.
4.
Abramavitz, M.Stegun, A .
/'
UTIAS Teehn1eal Nate No. 202
Institute far Aeraspaee Studies, Uni versi ty of Toranto
LAXERAL VIBRATIONS OF AN EXTENDING ROD
~
42 pages 70 f1gures
Ja.nkov1c, M. S.
1. Vibratlons 2. Canti lever beams
I. Jankovie, M. S. H. UTIAS Technieal Nate No. 202
The dynomies of an extending rad has been investigated in Refs. 1 & 2. The equation of lateral motion of a rod whose length changes wlth time has been deri ved there and was shown tO' be a fourth-order linear
partlat d.1fferential equo.tion, under the assunption ofaxial. rlg1d1ty and small defiections. The exact
selutton has not been found yet a.lthough a similarity Belutton has been found ror speciflc axial speed
history. Quasi-moda.l analysls was used to transfarm the partial. dif'ferentlal equation lnto an lnflnl ty of l.inear dlfferentlal equations with varlable coeff'iclents. One selutton was obtained in the farm of'
Bessel function for the first mode only.
One purpase of this wark is ta "ssess the contribution of the seeand rode ta the defleetian of a flexible rad during extension, and ta find an spproximate solutlon uslng the first and seeond rode whieh will be more appropri"te than the Bessel function approximation. As a prelude, tbe analysis of an extending
pendulum is done in Sectian H beeause of the sim111ll'ity with the case of an extending rad and because
of its simplicity, nallEly only one degree of freedom. This is equivalent in some sense to the ene-mode
representation in the rod case because 1 ts length changes in a s1m11ar menner to that of all extending rod. This allows certain techniques end concepts te be evBl.uated prior to the more challenging case which has an lnfinite number of degrees of freedom. There ls na exact Bolution available for the JendulUlll with arbltrary &tenslon rate, sO certa1n approximatlons are examined. The resultlng error ls
defined in nondimensional ferm, and plotted against nondimenslonal parameters to ensure generali ty of
the analysis. The spproximate analys is is then applied in Seetion III ta the object of real interest -the extending rad.
Available copies O~ th is report are limited. Return this card to UTIAS, if you require a copy.
UTIAS Techn1cal Nate No. 202
Institute for Aerospace Studies, University of Toronto
LAXERAL VIBRATIONS OF AN EXTENDING ROD
~
42 pages 70 figures
Jankovic, M. 3.
1. Vibrations 2. Cantilever beams
I. Jankovic, M. S. II. Ul'IAS Technical Nate No. 202
The dynamica of an extendlng rod bas been investigated in Refs. 1 & 2. Tbe equation of lateral motien of
a rod ",hase length changes with time has been derived there and was shawn te be a fourth-order linear
part1al differential equation, under the &B8U11ptian ofaxial rigidity and small defleetians. The exact
solution has not been faund yet althaugh a .1m11arity solutian has been found for speeific axial speed
history. QUasi-modal analysis was used to transform the partial differential equation inta an infinity of linear differential equatians with variabie eoefficients. One solution was obtained in the form of
Bessel function for the first made only.
One purpose of this wark ls to asseSB the contribution of the second mde to the deflection of a flexible
rod dw'ing extension, ond to flnd an approximate solutlon using the first and sccond oode whieh will be more appropriate than the Bessel .function approximation. As a prelude, the analysis of an extending pendulum 1s dO!1e in Sectlon II because of the s1.m11arlty with the case of an extending rod and because
of its simpllcity, namaly orü.y one degree of freedom. Thls is equivalent in some sense to the one-mode
representatlon in the rod case because i ts length changes in a similar manner to that of an extending rod. This allows certain techniques and concepts to be evaluated prior te the more challenging case
which has sn infinite munber of degrees of freedom. There is no exact solution available for the
pmdUlum \dtb arbitrary extension rate, so certain approximations are examined. The resUlting error is
defined in nondimensional form, and plotted against nandimensional parameters to ensure generality of
the ar.alysis. The spproxime.te analysis is then appl1ed in Section III to the object of real interest
-the extending rad.
Available copies of this report are limited~ Return this card to UTIAS, if you require a copy.
UTIAS Teehn1cal Note No. 202
In8ti tute for Aeraspaee studie8, Uni ver8i ty of Taronto
~
LAXERAL VIBRATIONS OF AN EXTENDmG ROD
42 pages 70 f'igures
Jankovic, M. S.
1. Vibrations 2. Cantllever beoms
I. Jankovie, M. S. H. UTIAS Technieal Note No. 202
The dynamics of an extending rad has been investigated in Refs. 1 & 2. The equation of lateral motion of
a rod whose length changes with time has been derived there and was shown to be a fourth-order linear partial d.1fferential equation, Wlder the assunptlon ofaxial rigld.1ty and small defiections. The exact
solution has not been found yet al.though a sim11ari ty solution bas been found :for specific axial. speed
history. QUasi-modal analysis was ufied to transform the partial. d.1fferential equation into an in1"lnl ty
of linear d.1fferential equations w1 th varlable coefflcients. One solution was obtained in the form of
Bessel function for the first mode only.
One purpose of thls work is to assess the contribution of the seeond rode to thc deflection of e. flexible
rod during extension, and to find an spproximate solution using the first and seeond rode whieh will be more approprie.te than the Bessel functian approximation. As a prelude, the analysis of an extending
pendulum is dane in Section H because of the s1m11arity with the case of an extending rad and because
of its simplicity, naJlEly' only one degree of freedom. This is equivalent in some sense to the one-mode
representation in the rod case because its leng th changes in a sim1lar menner to that of an extending roeL This allows certain techn1ques and concepts to be evaluated prior te the more challenging case
which has all infinite number of degrees of freedom. There is no exact solution available for the JendUlum with arbitrary extension rate, so certain approx1mations are exam1ned. The resulting error is
defined in nandimensianal form, and platted against nandimensional parameters· to ensure generali ty of
the analysis. The spproxime.te analysi8 is then applied in Seetian III to the object of real interest
-the extending rad. f h' h' A f
f\vallable coples 0 t IS report are limited. Return t IS card to UTI S, i you require a copy.
UTIAS Techn1cal Note No. 202
Insti tute for Aerospace Studies, Uni versi ty of' Toronto
~
LAXERAL VIBRATIONS OF AN EXTENDING ROD
42 pages 70 figures
Jankovic, M. S.
1. Vibration. 2. Canti lever beams
I. Jankovic, M. S. II. Ul'IAS Technical Note No. 202
The dynamics o:f all extending rod has been investigated in Ref's. 1 & 2. The equation of lateral motion 0: a rod whose length changes with time has been derived there and was shown to be a fourth-order linear
part1al differential equation, under the assu.;>tion afaxial rigidity end small deflections. The exact
solution has not been found yet although a similarity solution ha. been found for specific axial speed
hl.tory.· QUasi-modal analysis was used to transform the partial differential equation inta an infinity
of linear differential equatians with variabie eoeffieients. One solution was abtained in the form of
Bessol function for the first made anly.
One purpose of this work ls to assess the contribution of the second rode to the deflection of a flexible
rod during extension, and to find all approximate solution uslng the first and second mode which will be more appropriate than the Bessel function approximation. As a prelude, the analysis of all extending pendul:wrI is done in Section II because of the similarity with the case of en extending rod and because
of lts simpl.icity, namely only one degree of' f'reedom. This is equivalent in same sense to the one-mode
representation in the rod case because its length changes in a similar marmer to that of all extending
rod. This aJ.lows certain techniques and concepts to be evaluated prior te the more challenging case
which has all lnfinite number of degrees of freedom. There is na exact solution available for the
pendulum with arbitrary extension rate, 50 certain approximations are examined. The resUltlng error is
defined in nandimensianal form, and plotted against nondimensional parameters ta ensure generality of
the analysis. The approximate analysis is then sppl1ed in Sectian III ta the object of real interest
-the extending rad.