S. L.

Lau Nonlinear Vibrations of Pieoewise-

Senior Lecturer,
Mem. ASME.
Linear Systems by Incremental
W.-S. Zhang1
Research Assistant.
Harmonic Balance Method
The incremental harmonic balance (IHB) method is extended to analyze the periodic
Department of Civil and Structural
vibrations of systems with a general form ofpiecewise-linear stiffness characteristics.
Engineering,
An explicit formulation has been worked out. This development is of significance
Hong Kong Polytechnic, as many structural and mechanical systems of practical interest possess a piecewise-
Kowloon, Hong Kong linear stiffness. Typical examples show that the IHB method is very effective for
analyzing this kind of systems under steady-state vibrations.

I Introduction
Many structural systems of practical interest possess a building block approach for treating structural systems using
piecewise-linear stiffness such as structures with different ten- a describing function for the piecewise-linear elements.
sion and compression moduli, supports with motion-limiting In this paper, the incremental harmonic balance (IHB)
stops, and mechanical systems having clearances or snubbers, method is extended to solve systems with a general piecewise-
etc. Recently nonlinear elements have been widely used to linear stiffness. All formulae have been worked out in explicit
obtain better performance in vibration controls. Many of these form by using the step function which may yield higher com-
elements may be modeled as possessing piecewise-linear or putational efficiency.
piecewise-nonlinear characteristics. The IHB method was initially proposed in a series of papers
The free vibrations of simple piecewise-linear systems can for treating nonlinear periodic and quasi-periodic vibrations
be solved exactly, as introduced in some classical textbooks of continuous systems (e.g., Lau and Cheung (1981) and Lau
(e.g., Timoshenko, 1974). However, even for such a simple et al. (1982)). Recently, this method has been extended to
system, it is not easy to obtain a closed-form solution for a analyze the steady-state response of systems with dry friction
forced steady-state vibration (Galhond et al., 1987). by Pierre et al. (1985). This extension is of great significance,
As is well known, a common solution procedure for strongly as vibration control devices using dry friction are increasingly
nonlinear vibration problems is the numerical time integration. employed.
This approach can give both transient and steady-state re- Another frequency-domain analysis procedure for nonlinear
sponses for given initial conditions and specified frequencies. systems is the so-called Galerkin/Newton-Raphson (GNR)
However, there are some drawbacks to this approach. For method (see Ferri and Dowell, (1988), for example). In fact,
example, it is very expensive and difficult to obtain the fre- this method is exactly equivalent to the IHB method. Ferri
quency responses, especially in parameter regions where the (1986) has made a detailed comparison. The GNR method
response is multivalued; and it is impossible to obtain unstable differs from IHB only in the order in which the Galerkin (or
solutions which are definitely necessary for gaining an insight harmonic balance) method and the Newton-Raphson proce-
into the system behavior. dure are applied.
Another widely used approach is the harmonic balance (HB) The notable developments of analysis procedures made by
method. Maezawa et al. (1973, 1980) applied this method suc- Ling and Wu (Fast Galerkin method, 1987) and by Cameron
cessfully to symmetric and unsymmetric piecewise-linear single and Griffin (AFT method, 1989) essentially consist in a sub-
degree-of-freedom oscillators. Recently, Choi and Noah (1988) stitution of the harmonic balance step by a FFT algorithm.
analyzed a similar problem by this method using the FFT These developments can be considered as generalizations of
algorithm. Watanabe and Sato (1988) developed a nonlinear the IHB method. The authors share the same thought with
Wei and Pierre (1989), that it is essentially an efficient imple-
mentation of the IHB (or GNR) method.
According to our own experience, the use of FFT replacing
Permanent address: Research Institute of Engineering Mechanics, Dalian
University of Technology, Dalian, People's Republic of China.
the HB step can reduce the amount of algebraic manipulation
Contributed by the Applied Mechanics Division of THE AMERICAN SOCIETY and programming work. In general, it may be more efficient
OF MECHANICAL ENGINEERS for publication in the JOURNAL OF APPLIED M E - than the conventional IHB method, especially for those cases
CHANICS. in which the nonlinearity is rather complicated, and a consid-
Discussion of this paper should be addressed to the Technical Editor, Prof. erable number of harmonic terms is required in the solution.
Leon M. Keer, The Technological Institute, Northwestern University, Evanston,
IL 60208, and will be accepted until two months after final publication of the However, for the present piecewise-linear problems as well as
paper itself in the JOURNAL OF APPLIED MECHANICS. Manuscript received by the for most practical structural vibrations, as long as explicit
ASME Applied Mechanics Division, May 2, 1989; final revision, July 27, 1990. formulae for integrals arisen in the Galerkin procedure are

Journal of Applied Mechanics MARCH 1992, Vol. 5 9 / 1 5 3

Copyright © 1992 by ASME
Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]
 Q= - and f=- (7)
•sfkTm " 2\jmk
are the frequency ratio and the damping ratio, respectively.
I l l I H B Analysis
Since the IHB method has been well expounded in a number
of papers (Lau and Cheung (1981), Lau et al. (1982), Pierre
et al. (1985), Pierre and Dowell (1985), for example), the re-
quired formulation will be briefly stated here.
Assuming that the current state of vibration corresponding
Fig. 1 Restoring force versus displacement relationship of the general to the excitation parameters Q0 and q0 is denoted by zo (6), a
piecewise-linear spring neighboring state can be expressed by adding the corresponding
increments as follows:
available or only a small number of harmonic terms is sufficient
to give satisfactory results, the conventional IHB method is Z(0) = zo(0) + Az(0)
still more effective than that of using FFT algorithm. q(6)=q0(d)+Aq(d)
In the present work, the extended IHB method is applied to
single degree-of-freedom (DOF) systems to demonstrate its fl = O0 + AQ. (8)
effectiveness. Two examples of practical significance in some For a small increment Az, the function G(z) may be written
mechanical devices are considered. The convergence of the formally as
solution sought is examined by increasing the number of har-
monic terms. It is shown that only a few harmonic terms are G(z)=G(z0) + G'(z0)Az (9)
sufficient to obtain satisfactory results. where G' (z0) represents the value of the first derivative with
The technique presented in this paper can easily be extended respect to z at z 0 which can be expressed with the step function.
to multiple DOF systems with piecewise-linear stiffness and Following the procedure in Lau et al. (1982), a linearized in-
piecewise-linear damping. cremental equation is obtained in the form

II Equation of Motion ti20d~^ + 2{Q0^ + Az + jG'(z)Az = R + AQF+AQ (10)

Consider the single DOF system consisting of a mass, a
viscous damper, and a piecewise-linear spring of general form. where
The equation of motion can be written as
R= - n2 d2Zo , ». n dzo , , 1 ^,, . 1
(ID
cfz dz ...
m ^ + cJt+f{z)=q(t) (1)
F--2Q0—r-2^- (12)
where
f{z)=kz + G{z) (2) AQ = Aq/k. (13)
and m, c, k, t, and z denote the mass, viscous damping coef- R is the corrective term and F is the unbalanced force term
ficient, linear spring constant, time, and displacement, re- due to unit frequency shifting.
spectively. The nonlinear restoring force G(z), which has a An approximate periodic solution may be expressed as
general piecewise-linear characteristic, is shown in Fig. 1 and N N
Zo = Zo
can be expressed as follows: 2 " ~ a°^ + 2 ("" C0S H® + bn s m nO) (14)
G(0) = 0, e0 = 0, e_p=-oo, e„=oo
and, accordingly,
GU)=A"_i(z-e_(i_1))+G(e_(f_i)), e_;<z<e_ ( i_i) N N
(/=l,2,...,p) Az= 2 ] Az„ = Agp/2 + y ] (Aa„ cos nd + Ab„ sin nd) (15)
n=0 n=\
G(z)=Ki(z-ei-1)+G(ei-i), e,_i<z<e,-
(i=l,2,...,q) (3) where N = the number of harmonic terms taken.
The Galerkin procedure is carried out with the Aa„'s and
where the k + K, (< = -p, - (p - 1), . . . , - 2 , - 1 , 1, 2, Ab„'s as the generalized coordinates:
. . . , (q - 1), q) are spring constants (Fig. 1), and I A: I is „2T
assumed to be greater than zero. d2Az dAz 1
The exciting force q(t) can be represented by a general d&5- + 2 f f l 0 dd
^ r + Az + 7 G'(z)Az S(Az) dd

-r
periodic function:
N N
q
(R + AQF+AQ) 6(Az) dd. (16)
1 (0 = 2 " =f°/2 +
2 Un cos nut + g„ sin nwt) (4)
n=0 «=0
Denoting the amplitude vector and its incremental vector by
where/o,/„, and g„ {n = 1, 2, .. iV) are force component
amplitudes of the corresponding harmonic terms. a=[a0,al..aN,bi,b2..bN]T (17)
For convenience, we introduce a phase angle 8 given by Aa= [Aa0,Aal..AaN,AbuAb2..AbN]T, (18)
) = ut (5) respectively, a system of 2N + 1 linearized equations in terms
of the increments Aa„'s and Ab„'s is obtained from Eq. (16)
and change the independent variable to 6 from t. Equation (1)
in the form of
is transformed into
CAa = R + AQF + AQ (19)
n2 +2 + G
§ ^f ^ ^=> (6)
C=
"[C,] [C12]
[C21] [C2]
(20)
where

154/Vol. 59, MARCH 1992 Transactions of the ASME

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 z z
a- f21
[C 1 2 ]f = -r G'(zo)cosid sinjd de (35)
/ /c J 0
/ a- f2,r
/
[C2i]f- = -r G'{zo)sinid cosjd dd (36)
A: J 0
e
o 1 f(z) I V 9
M-2 e
M-1
1 r2T
SM,
% e
i e2 "\; ; i j7 T
A-i
[C2]f" = - j G'(z0)smi8 sinje dO (37)
e
/ -2
D/VZ. a,- f2'
= G( z cos e de
~TJ - °) > (38)
2TT
Fig. 2 Zeros of equations z = e„(n = - p , - p + 1,..., - 2 , - 1 , 0 , 1 , i r27r
R L
2 <7 - 1, q) 2/ =-^j G(z0)smiddd (39)

where G' (z0) can be expressed by the use of the step function.
The evaluation of the integrals (34)-(39) requires a knowl-
Ri edge of the zeros of the equations z = en (n = - (p - 1),
R= (21)
. . . , - 2 , - 1, 0, 1,2,. . . , q — 1). In the computer program,
this is achieved at each iteration through a procedure which
"Fi" uses bisection and interpolation methods on the trigonometric
(22)
A equation z = e„. Let 8U 62, . . . , 8M (f>\ < 82 < • . . < 8M)
be the M zeros of z = e-.{p.x), z = e~ip-2), . . . , z = e_i,
AQ,
AQ = (23) z = e0, z = eu . . . , z = e 9 _! and su„, s2,„, . . . , sM+l,„ be
AQ2
the value of the sign function of z - en (n = —p - 1,
The explicit expressions for C, R, F, and AQ have been 0, 1, . . . , q), respectively, in the intervals [80, 6{\ . . . [8M,
worked out as follows: 8M+ 1], with d0 = 0 and 8M+, = 27r, and denote the step function
(1) Elements of the matrix C:
[C,]j/ = a y 6 < ^ r ( l - / ^ ) + [ C 1 ] ^ ,
(i = 0,l,...,NJ = 0,l,...,N) (24)
[C12]i/ = 25,y7rfy'Q0+[C12]f> The integrals (34)-(39) can then be expressed explicitly in
(i = 0,l W,y=l A0 (25) terms of 8U and S„,„:
M r -1
[c 21 ]y=-25, yT ryo 0 +[c 21 ]f ) [Ci]J' L = - 7 J 2 J I 2 J ^ ( J » t i , i A t i , i + i ) ^ ( ' 4 i j ( 8 « t i )
(i=l,2,-,NJ = 0,\,...,N) (26) u = 0 vu = - p

[ 0 2 ^ = 5 ^ 1 - y "QS)+ [cy^
('=1,2 N,j=l N) (27)
where
(41)
„a _ .1 forn^O M
« " 1/2 f o r « = 0 „. f -'
[Cl 2 ]y = - T 2 j j ^ J ^ ^ U + l . l " • S K+l,U+l)^l'(-Sv(^K+l)
5y is the Kronecker's notation. W = 0 U = -/J

(2) Elements of vectors R, F, and AQ: -5, y (e„)) + 2 ^ ( ^ + i , ^ i A + i , « ) ^ ( 5 y ( ^ + i ) - ^ ( ^ ) )

B=I J
2
R i / = - a , - (l-/ Q§)a, + 2r/fioft,-| ir + K],- , (42)
(/ = 0,1 N) (28) =
[C2i],y T.2jj Z_j H{su+iv,su+\,v+{)Kv{Cjj(du+\)
2 « = 0 v y = -/>
R2,= (l-i Q§)6,-2frQ 0 B / -^ TT + R 2 / ,
1
(i=l,2,...,N) (29) „)/('„( Cy(6„+ l) ~ Cjj(dU))
V=\
F 1/ = 2ir/(/Ob«/-f*/). (/=0,1,...,A0 (30)
(43)
F 2/ = 2«(iOo6,+ ffl,)I (i=l,2,...,A0 (31)
j A/ (- - 1

AQu = T onwAfi, (/ = 0,1,...,N) (32)' [C 2 ]f = - 2 2 ^(^+i.«A+i.»+i)^«(^( e «+i)

u = 0 v y = -/?

AQ 2 , = - 7 r A g ; , ( / = 1 , . . . , / V ) . (33)
u=l
The nonlinear part of the elements of C and of the components
of R are expressed by: (44)
M C -1

J 0
2 7T

G'(z0)cosi8 cosjd dd (34)

™i; —" , 2_j j 2-i
u = 0 v y = ~p
H(su+i,i"su+i,v+i)

Journal of Applied Mechanics MARCH 1992, Vol. 5 9 / 1 5 5

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 N
X [Kv J ] aj[aj[Au(Bu^)-Au(eu)-\
I
>=0

+ bJ[Bij{du+1)-Bu{el,)]] + Kv[Ei(eu+l)-Ei(eu)]
N
aj[Ay(6u+i)

+ U£/(0 B + i)-£'/(fl«)])[ (45)

JVL= 1
R•2i i- Z j j 2-J H(Su+l,t»su+l,v+l)
u = 0 k u = -/?

N Fig. 3 A single DOF system with clearance, m = 1, k = 1, e_, - -0.2,

/ e, = 0.2, K, = /f_, = -k, K2 = K_2 = 0, f = c/V^TSS = 0.01, q(t) =
x (*„ 2 uJ[aj[Cv(Ou+l)-CuiOu)]
COSut

frequency increment is prescribed and Afl is kept to be zero

+ ^[A7(e„ + 1 )- J D/ / (e„)]] + /c„[F,(0„ +1 )-F,.(e„)] in the subsequent iteration process. At every increment, one
has to iterate with the following equations,
9 / N
C (/) Aa (/+D = R <o (54)

l)=l \ y=0 ,('+!)_ a (,) + Aa (i'+i) (55)

- Q ( 0 J ] + &,[£>(,(*,,+ , ) - Z f y t f J ] ] and the updated matrices C' , R +
u 1 JJO'+D
( , + 1)
are then evaluated from
Eqs. (24)-(29)intermsofthe (/ + l)th amplitude vector a (,+ 1).
+* B [/ r /(«« + l)-^(»«)] (46) The other option is by ^-incrementation; i.e., a properly
chosen amplitude component increment Aak or Abk (usually
where
the one that is in resonance and varies faster) is prescribed,
K_r=(Js:_1-^_2)e_1 + (Js:_2-/:_3)e_2 + ... and in the subsequent iteration process this Aak (or Abk) is
kept equal to zero. In this case, one has to iterate with the
+ (^. + i - ^ _ f ) e _ , + 1 following equations,
K r =(^,-if 2 )e 1 +(A: 2 -X3)e2+... + ( ^ - i - ^ ) e ^ 1 (47) C u(/), Aa
A,(/+!)_
R ( , ) +AO (<"+i)ir<i) (56)
( / + i ) . a (,) + Aa (i+D
sia(i—j)9 sin(/+y')0 a (57)
Ay(e)=- (48)
U-J)0 (i+j)0 fi<'+1» = Q('»+AQ (/+1) . (58)
(,+1 (,+1)
The updated matrices C \R are evaluated from Eqs.
cos(i—j)6 cos(/+y')0 (24)-(29) in terms of the (/ + l)th vector a (/+1) and Q (,+ 1) for
a-j)e U+J)6
(49)
the next iteration.
Bij(6) •• Depending on the local relative slope of the response curve,
cos2/0 one of the incrementation options is used. To obtain the next
U=j)
2/0 point of solution the iteration process proceeds until the cri-
terion
cum=BJI(e) (50)
Aa,-
sin (i-j)d sin (i+j)6 max <e (59)
Du(6)=- (51) «ref
(i-j)0 U+j)0
is satisfied, where e is the permissible error depending on the
sin/6 accuracy required, Aa,- represents the rth component of the
E,(0)=0 (52)
If most recently computed increment, and aKs stands for the
cosid largest component of the vector a.
(53) In general, the number of iterative steps required for con-
id vergence is problem dependent. Nevertheless, the required
In expressions (48)-(53), the limiting values should be taken number can be reduced by using an extrapolation procedure
if both the numerator and denominator become zero. to obtain a prediction for the next solution point from the
known ones before the iteration process is started.
IV Numerical Examples The required number of iterations to converge is obviously
To obtain a frequency response curve of a system, an initial dependent on the magnitude of the increments in an incre-
state should first be given. This state need not be a solution, mental step as well. The smaller the increment, the fewer the
but an assumed vibration state will do as long as the initiali- iterative steps required. In our computer program, the desired
zation process can converge to a required solution (e.g., super/ number of iterative steps to converge can be set by the user,
subresonance, fundamental resonance, etc.). The incremen- and the program will adjust automatically the arc length of
tation procedure for tracing the frequency responses (usually the extrapolated increment step by step, so as to retain the
one of the their branches) for a constant force level is similar preset number of iterations as close as possible during the
to that described in Lau et al. (1982). There are two options incrementation process. In the following examples, the desired
in this procedure. One is the ^-incrementation option; i.e., the number of iteration for convergence is set to be four.

156 / V o l . 59, MARCH 1992 Transactions of the ASME

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 0.06

Fig. 4 Maximum amplitude versus frequency; curve 1: free vibration

backbone; curve 2: superharmonic/fundamental response; curve 3: sub-
harmonic response. A = (zmax - zmin)/2.

Fig. 5 Maximum amplitude versus frequency; curve 1: shifted free vi-

bration backbone; curve 2: superharmonic/fundamental response; curve
3: subharmonic response. A = (zmax - zm,„)/2.

Example 1: Single DOF System With Clearance. The free vibration backbone curve by multiplying the frequency
mathematical model of this problem and its related parameters by three, it is found that the subharmonic resonance response
are shown in Fig. 3. In the analysis, two odd sine terms and is located in the neighborhood of the shifted backbone curve
cosine terms are adopted. The computed maximum amplitude (see Fig. 5, curve 1'). A similar phenomenon is observed for
versus frequency responses are shown in Fig. 4. Curve 1 rep- the superharmonic resonance by shifting the backbone curve
resents the free vibration backbone curve; the frequency ap- to the left side with the frequency divided by three. These
proaches 1.0 with increase of amplitude and reduces to zero interesting facts reveal the free vibration response, which re-
when the amplitude decreases at the clearance boundary. Al- lates to the inherent characteristics of the system, affects the
though the clearance does not cause much effect on the fun- behavior of various kinds of resonances.
damental resonance (curve 2), this kind of nonlinearity gives In the case of no damping, the subharmonic response extends
rise to superharmonic and subharmonic resonances (see curves to infinity at both positive (in-phase solution) and negative
2 and 3, respectively). However, the existence of damping, (out-of-phase solution) branches. The sub-harmonic response
even when this is of a rather small magnitude, causes both the shrinks into complicated closed curves in the presence of damp-
super and subharmonic resonances to be greatly suppressed, ing as shown in Fig. 6, where am, bi/3, au and bu are the
in comparison with the fundamental resonance. If we shift the amplitudes for the cos(atf/3), sin(co?/3), cos(utf) and sin(u>0

Journal of Applied Mechanics MARCH 1992, Vol. 5 9 / 1 5 7

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 Fig. 6(a) Amplitudes am, t>„3 versus frequency 0

T- - 0 . 1

Fig. 6(6) Amplitudes a„ 6,, versus frequency a

Fig. 6 Subharmonic response amplitudes, z(l) = a1/3 cos(wf/3) + b„

sin(wf/3) + a, cos(uf) + t>i sin(wf)

terms, respectively. In either case, the subharmonic response f(z)

is separated from the superharmonic/fundamental response
solution. k+K 2
It should be noted that in order to obtain the third-order
subharmonic resonance without modifying the above formu-
lation, the exciting force q(t) in Eq. (4) should be taken to k+Ki
be q(t) = cos 3 wt. After the computation, the subharmonic
response is obtained by shifting with the frequency divided by e-1
three. For the separate subharmonic resonance, a starting ei
guessed state may be obtained by referring to the corresponding k+K-i
shifted free vibration backbone curve, such as the one shown
in Fig. 5.
k + K-2
Example 2: Single DOF System With Symmetric Piecewise-
Linear Stiffness. The mathematical model and its parameters
are shown in Fig. 7. The computed results with two odd sine
terms and cosine terms under a single harmonic excitation are Fig. 7 A single DOF system with symmetric piecewise stiffness, m :
plotted in Fig. 8. The third harmonic response curves appear 1, k = 1, K, = K_, = 0, 2 = K_2 = 1, e_, = - 0 . 5 , e, = 0.5, f ••
to be quite complicated. This is basically due to a rotation of c/V4m/t = 0.01, q(t) = cosoit

158/Vol. 59, MARCH 1992 Transactions of the ASME

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 Fig. 8(a) Amplitudes au i>, versus frequency 11

£X

Fig. 8(b) Amplitudes a3, b3 versus frequency SI

Fig. 8 Fundamental responses of symmetric piecewise linear system,
z(t) = a,coswf + b,siiW + a3cos3iof + o3sin3ur

more than 2ir in the phase angle. The subharmonic resonance Table 1 Time histories for to = 1 obtained with different
is also separate under the present excitation level, and has a number of harmonic terms (NH)
similar response to that of example 1. When K2, K_2 are large NH 6 0 TT/6 TT/3 TT/2 2TT/3 5x/6 ir
in comparison with k, this model represents a system having
a massless snubber which is widely used in some vibration 3 5.352 4.813 2.980 0.534 - 2.048 -4.176 -5.352
5 5.359 4.793 2.985 0.529 -2.041 -4.172 -5.359
control devices. Figure 9 shows the total amplitude response 8 5.360 4.793 2.985 0.529 - 2.043 -4.173 -5.360
of such a system in which the snubber stiffness is ten times
that of the spring stiffness. The results are obtained with three Table 2 Time histories for co = 2 obtained with different
odd sine and cosine terms. It can be seen that for a given n u m b e r of h a r m o n i c terms (NH)
frequency, the amplitude is much less than that of the linear NH 6 0 w/6 TT/3 TT/2 2TT/3 5 TT/6 •w
counterpart. Note that the right portion of the response curve 3 8.607 9.330 6.971 3.018 -1.160 -5.303 -8.607
above ,4 = 5 represents unstable solutions. 5 8.613 9.325 6.974 3.022 -1.169 -5.293 -8.613
To show the convergence of the obtained solutions, this 8 8.612 9.325 6.975 3.024 -1.168 -5.291 -8.612
problem has been also computed with five and again eight
harmonic terms (i.e., up to cos(15co?) and sin(15co?)). The time monic responses, the corresponding harmonic terms should be
histories (z versus 6) obtained for co = 1 and 2 are listed in included.
Table 1 and Table 2, respectively. We can see that very good
results are obtained by using only three harmonic terms for V Concluding Remarks
these cases. Of course, to analyze higher-order super/subhar- The IHB method has been extended to the analysis of systems

Journal of Applied Mechanics MARCH 1992, Vol. 59/159

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]

 < 10

Fig. 9 Maximum amplitude versus frequency for the system shown in

Fig. 7 with parameters: m = 1, k = 1, K, = /(_, = 0, K2 = K_2 = 9,
e_, = - 5 , e, = 5, f = 0.01, q = coswf. A = (zmax - zm[„)!2.

with a piecewise-linear stiffness of general characteristics. All Ferri, A. A., 1986, "On the Equivalence of the Incremental Harmonic Balance
Method and the Harmonic Balance—Newton Raphson Method," ASME JOUR-
formulae have been worked out in explicit form. This method NAL OF APPLIED MECHANICS, Vol. 53, pp. 455-456.
can be further extended to treat multiple-DOF systems with Ferri, A. A., and Dowell, E. H., 1988, "Frequency domain solutions to multi-
piecewise-linear stiffness as well as piecewise-linear damping. degree-of-freedom, dry friction damped systems," / . Sound and Vibration, Vol.
Moreover, the present technique can be applied to the so-called 124, No. 2, pp. 207-224.
Friedmann, P., Hammond, C. E., and Woo, T. H., 1977, "Efficient numerical
ramp function used in modeling dry friction dampers (Ferri treatment of periodic systems with application to stability problems," Int. J.
and Dowell, 1988). The explicit formulation used here yields Num. Math. Engg., Vol. 11, pp. 1117-1136.
higher computational efficiency. Galhoud, L. E., Masri, S. F., and Anderson, J. C , 1987, "Transfer Function
In contrast to the time-step integration method, both the of a Class of Nonlinear Multidegree of Freedom Oscillators," ASME JOURNAL
OF APPLIED MECHANICS, Vol. 54, pp. 215-225.
stable and the unstable vibration states are obtained by the Hsu, C. S., 1972, "Impulsive Parametric Excitation: Theory," ASME JOUR-
IHB method. The stability analysis can be performed by the NAL OF APPLIED MECHANICS, Vol. 39, pp. 551-558.
use of Hsu's method (Hsu, 1972, 1973, Friedmann, 1977) on Hsu, C. S., 1973, "Applications of the theory of impulsive parametric ex-
the corresponding variational equations. It is worth pointing citation and new treatments of general parametric excitation problems," ASME
JOURNAL OF APPLIED MECHANICS, Vol. 40, pp. 78-86.
out that the variational equations are identical to the incre- Lau, S. L., and Cheung, Y. K., 1981, "Amplitude Incremental Variational
mental equations (with the right-hand side term set to zero) Principal for Nonlinear Vibration of Elastic System," ASME JOURNAL OF AP-
used in the IHB method. PLIED MECHANICS, Vol. 48, pp. 959-964.
The subharmonic resonances are closely related to the oc- Lau, S. L., Cheung, Y, K., and Wu, S. Y., 1982, "Variable Parameter
Incrementation Method for Dynamic Instability of Linear and Nonlinear Sys-
currence of chaos, which has become a prominent topic over tem," ASME JOURNAL OF APPLIED MECHANICS, Vol. 49, pp. 849-853.
the past few years in engineering and applied science. Some Lau, S. L., Cheung, Y. K., and Wu, S. Y., 1983, "Incremental Harmonic
interesting problems involving piecewise-linear systems with a Balance Method With Multiple Time Scales for Nonlinear Aperiodic Vibra-
single change in stiffness have been dealt with by Thompson tions," ASME JOURNAL OF APPLIED MECHANICS, Vol. 50, pp. 871-876.
Lau, S. L., Cheung, Y. K., and Wu, S. Y., 1984, "Nonlinear Vibration of
(1983) and Shaw and Holmes (1983) associated with the ex- Thin Elastic Plates: Part II Internal Resonance," ASME JOURNAL OF APPLIED
istence of chaotic motions. The IHB method can be applied MECHANICS, Vol. 51, pp. 845-851.
to systems of this kind for seeking subharmonic resonances. Ling, F. H., and Wu, X. X., 1987, "Fast Galerkin method and its application
Recently, Dowell and Pierre (1984) in a review paper sug- to determine periodic solutions of non-linear oscillators," Int. J. Nonlinear
Mechanics, Vol. 22, pp. 89-98.
gested the IHB method could be used for studying the self- Maezawa, S., and Furukawa, S., 1973, "Superharmonic resonance in
excited, chaotic oscillation of autonomous systems. We are piecewise-linear system," Bulletion of the JSME, Vol. 16, pp. 931-941.
looking forward to seeing such a development in the near Maezawa, S., Kumano, H., and Minakuchi, Y., 1980, "Forced vibrations in
future. an unsymmetric piecewise-linear system excited by general periodic force func-
tions," Bulletin of the JSME, Vol. 23, pp. 68-75.
Pierre, C , and Dowell, E. H., 1985, " A Study of Dynamic Instability of
Acknowledgment Plates by an Extended Incremental Harmonic Balance Method," ASME JOUR-
This work has been supported by a Research Grant from NAL OF APPLIED MECHANICS, Vol. 52, pp. 693-697.
Pierre, C , Ferri, A. A., and Dowell, E. H., 1985, "Multi-harmonic Analysis
the Hong Kong Polytechnic. The authors wish to thank Prof. of Dry Friction Damped Systems Using an Incremental Harmonic Balance
M. Anson and the reviewers for their comments which were Method," ASME JOURNAL OF APPLIED MECHANICS, Vol. 52, pp. 958-964.
very helpful in the revision of this paper. Shaw, S. W., and Holmes, P. J., 1983, " A periodically forced piecewise linear
oscillator," J. Sound and Vibration, Vol. 90, No. 1, pp. 129-155.
References Thompson, J. M. T., 1983, "Complex dynamics of compliant offshore struc-
Choi, Y.S., and Noah, S.T., 1988, "Forced periodic vibration of unsymmetric tures," Proc. R. Soc, London, Vol. A387, pp. 407-427.
piecewise-linear systems," / . Sound and Vibration, Vol. 121, No. 1, pp. 117- Timoshenko, S. P., Young, D. H., and Weaver, W., Jr., 1974, Vibration
126. Problems in Engineering, 4th ed., John Wiley and Sons, New York.
Cameron, T. M., and Griffin, J. H., 1989, "An Alternating Frequency/Time Watanabe, K., and Sato, H., 1988, "Development of Nonlinear Building
Domain Method for Calculating the Steady-State Response of Nonlinear Dy- Block Approach," ASME Journal of Vibration, Acoustics, Stress, Reliability
namic Systems," ASME JOURNAL OF APPLIED MECHANICS, Vol. 56, pp. 149- in Design, Vol. 110, pp. 36-41.
154. Wei, S. T., and Pierre, C , 1989, "Effects of dry friction damping on the
Dowell, E. H., and Pierre, C , 1984, "Chaotic oscillations in mechanical occurrence of localized forced vibrations in nearly cyclic structures," J. Sound
systems," Chaos in Nonlinear Dynamic Systems, J. Chandra, ed., SIAM. and Vibration, Vol. 129, No. 3, pp. 397-416.

1 6 0 / V o l . 59, MARCH 1992 Transactions of the ASME

Downloaded From: [Link] on 12/25/2015 Terms of Use: [Link]