This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

**Bifurcation in the dynamical system with clearances
**

N. Kranjcevic* University of Zagreb Zagreb, Croatia M. Stegic† University of Zagreb Zagreb, Croatia N. Vrankovic‡ University of Zagreb Zagreb, Croatia

Abstract— Considering the time finite element method and the piecewise full decoupling method, the bifurcation in the harmonically excited two-degree-of-freedom dynamical system with clearances is investigated. The stability analysis based on the utilization of the Poincaré map is used in the time finite element procedure. In the piecewise full decoupling computations, the bifurcation analysis is performed by studying the periodicity of steady state solutions. Close agreement is found between the obtained results. Keywords: nonlinear vibrations, clearances, bifurcation

vectors of mean and alternating load, respectively. Furthermore, the damping and stiffness matrices Ζ and Ω are given as:

2 ⎤ ⎡ 1 ⎡ ζ 11 −ζ 12ω12 ⎤ −ω12 ⎥. ⎥, Ω =⎢ Ζ = 2⎢ 2 ⎥ ⎢ 2 ⎢ ⎥ ⎣−ζ 21ω21 ζ 22ω22 ⎦ ⎣−ω21 ω22 ⎦

(2)

The nonlinear displacement vector is defined with the piecewise linear functions h(qi ) , i=1,2 that describe the clearance of value 2 (figure 1):

⎧qi + 1 ⎪ ⎪ ⎪ h( qi ) = ⎪ ⎨ 0 ⎪ ⎪ ⎪ qi −1 ⎪ ⎩ qi < −1 −1 ≤ qi ≤ 1 . qi > 1

I. Introduction Nonlinearities such as gaps and clearances exist in many dynamical systems either by design or due to manufacturing tolerances and wear. The characteristics of systems with clearances include abrupt variation of stiffness which is inherently nonlinear and difficulties that occur in their analysis exhibit great diversity. In this paper, the bifurcation in the two-degree-offreedom dynamical system with clearances, subjected to periodic excitations, is studied. The bifurcation analysis is performed considering the time finite element method and the piecewise full decoupling method. The system is assumed to be controlled by the excitation frequency (codimension-one bifurcation). A precise value of the excitation frequency, at which the bifurcation occurs is determined. II. Problem formulation The mechanical model of a two-degree-of-freedom system with clearances consists of two mass elements, two linear viscous dampers and two clearance type nonlinearities. The system is periodically excited. With q T = [q1 , q2 ] being the nondimensional displacement, the equation of motion can be written as:

q ′′ + Ζ q ′ + Ω h(q ) = f 0 + f a cos(ητ )

(3)

h(qi)

1

-1 1

1

qi

Fig. 1. Piecewise linear displacement function

(1)

where hT (q ) = [ h(q1 ), h(q2 ) ] is the nonlinear displacement vector, η denotes a nondimensional excitation frequency while f 0 and f a are the amplitude ______________________

* E-mail: nenad.kranjcevic@fsb.hr † E-mail: mstegic@fsb.hr ‡ E-mail: nvankov@fsb.hr

Exact solutions of piecewise linear equations of motion are very rare and almost all of the solution methods are only approximate. A common solution method is the numerical time integration (Runge Kutta etc.), but this procedure can be very time consuming particular inside the frequency range of multi-valued responses and for lightly damped systems, since a great number of excitation periods must be simulated to obtain a steady state response. A very efficient method for solving piecewise linear equations of motion in the frequency domain is the harmonic balance method [1], [2]. When the assumption of dominance of the first harmonic in the response is satisfied, the harmonic balance method is very accurate and numerically reliable procedure. If the influence of higher harmonic responses is significant, the method becomes very unreliable [13]. The incremental harmonic balance method [14] provides the study of effects of superharmonics and subharmonic to response and may be successfully applied to wide range of nonlinear diferential equations. This method is very

2007 convenient for determining frequency response characteristics. because a new solution can be sought. [6] is a new developed numerical procedure of explicit integration based on coupling series of linear solutions. the system moves away from its stable periodic behavior as soon as a small perturbation is superimposed. with a summation notation representing the element assemblage operation: q = ∑ q j . the stiffness discontinuity in clearances can produce qualitative changes in the features of system. j =1(1)ne using the standard finite element (6) where Nj is the shape function and q j denotes the nodal displacement. j =1 j =1 ne ne τ2 τ1 (4) Lagrangian and the respectively. Similar to the standard finite element technique. (5) For a periodic solution. called bifurcations. ′ q′ j (τ )=N j (τ ) q j . can be expanded into Taylor series retaining only the linear . Considering Newton-Raphson method. the can be written as: τ2 ∫ {δq′ q′ + δq τ T 1 T [ − Ζ q ′ − Ω h(q )+f 0 +f a cos(ητ )]} dτ . the nodal displacement q is iterated as: = δq p T τ2 τ1 q (n+1) = q (n ) + ∆q (n ) (11) The basis of the time finite element procedure is. A number of harmonics used in the solution is limited by numerical efficiency of the method. The global finite element equation is a set of nonlinear algebraic equations in unknown nodal displacement vector q which can be determined employing one of iteration methods. as the periodic boundary condition requires both displacements and generalized momenta to be identical at τ 1 and τ 2 . The above interpolation inserted into the integral form (5) at time element level results in the local finite element equation: A j q j + g j (q j )+f j = p j . The piecewise full decoupling method [4]. the division of the time interval τ 2 − τ 1 (equals to the period For a small increment ∆q (n ) . The bifurcation is characterized by the fact that beyond the critical excitation frequency. The global element equation can be determined by assembling the finite element equations at the time element level (7). while p generalized momenta. The accuracy of the method does not significantly depend on a magnitude of the integration step. Hamilton principle (4) g (q ) = ∑ g j (q j ). A complete characterization of dynamical systems with clearances particularly requires determination of the stability of their steady state solutions. j =1 ne f = ∑ fj j =1 ne . The solution for all the spatial degrees of freedom at all time steps within a given time interval is sought through a set of algebraic equations.12th IFToMM World Congress. Besançon (France). τj τ j +1 (8) fj = ∫ [N τj T j f 0 +N jT f a cos(ητ )]dτ . as in the standard finite element modeling scheme. since the computer time grows geometrically with the number of harmonics in assumed solution. yielding: Aq + g (q )+ f =0 (9) where. with the previous solution used as very good approximation. since the results inside the linear part are based on analytical solutions. It is well known that under certain periodic excitations. June18-21. the equation (9). the term on the right-side of the equation (9) vanishes. The time finite element method [3]. III. [12] is an implicit method for prediction of a steady state response of the nonlinear dynamical systems. associated with piecewise linear functions. τj τ j+1 g j (q j ) = − ∫ N jT Ω h(q ) dτ . (10) where L and δW are the nonconservative virtual work. j =1(1)ne (7) where: τ j+1 Aj = ∫ [N ′ j T T ′ N′ j − N j Ζ N j ] dτ . denotes the vector of the Employing the equation (1). Fundamentals of the time finite element method The time finite element is based on Hamilton weak principle which describes the motion of the mechanical system between the two known nondimensional times τ 1 and τ 2 : τ2 τ1 T ∫ (δL + δW ) dτ = δq p of excitation) into a finite number ne of time elements and the approximation of the displacement q j (τ ) within each time element interpolation: q j (τ )=N j (τ ) q j . the time interval is divided into a finite number of time elements. A = ∑ Aj .

cyclic-fold or period-doubling). Poincaré map P is defined as: P (s ) = x (s . resulting in a generalized static bifurcation (transcritical. K (n ) (n ) ∂g (n ) (q ) . ⎡ 1 Π 2Ω = Π8Ω = ⎢ 2 ⎣ −ω21 0⎤ . p ] T where B and I correspond to the temporal nodes at the ends and in the interior of time element j. depend on the piecewise linear displacement function (3).12th IFToMM World Congress. In a neighborhood of the fixed point. This procedure results into the global equation: ⎡ d1 ⎤ ⎡ K11 ⎢ ⎥+⎢ ⎢ d 2 ⎥ ⎢ K 21 ⎣ ⎦ ⎣ K12 ⎤ ⎡ ∆q1 ⎤ ⎡− p1 ⎤ ⎥⎢ ⎥=⎢ ⎥ K 22 ⎥⎦ ⎢⎣ ∆q2 ⎥⎦ ⎢⎣ p2 ⎥⎦ (20) i. There are two possible scenarios in which the eigenvalues may leave the unit circle.τ 2 ) = s Considering the assemblage procedure. k = 1(1)9 (22) where: Π1Ω = Π 3 Ω = Π 7 Ω = Π 9 Ω = Ω . To incorporate determination of the eigenvalues of Poincaré map in the time finite element procedure. Π5Ω = 0 2 ⎥ ⎣ 0 ω22 ⎦ denote the local stiffness matrices while: T 2 2 2 ⎡ + ω22 b1T (−ω21 )⎤ Ω = −b9 Ω = ⎣ (1 − ω12 ). The iteration procedure defined by the equations (11). Second. Besançon (France). Compare the lower part with the upper part of the equation (18) to eliminate the interior nodal displacement ∆qI j : −1 (d B − K BI K II dI ) j + −1 (K BB − K BI K II K IB ) j ∆ qB j = pB j . yielding the elimination of the common node. in equations (14). −1 ⎥ − K 22 K12 ⎦ (21) (16) where v is the norm of the deviation from the fixed IV. we have: ∂P P (s + v ) = P (s ) + v + O( v 2 ) ∂s where 1 and 2 denote the initial and final time nodes. The stability of time finite element solutions can be examined as the stability of the Poincaré map. we have to return to the incremental finite element equation (12) at the time element level: The piecewise full decoupling method is based on the substitution of the piecewise linear equation of motion (1) with a set of linear equations of motion. (12) will be terminated when the increment ∆q (n ) converges towards zero. ⎦. the equation (19) of all time elements are further condensed by imposing the compatibility conditions at the common node of the two elements. = A+ ∂q (n ) (n ) (13) ⎡ ∆q B ⎤ ⎡p ⎤ ⎢ ⎥ = ⎢ B⎥ ⎢ ∆q I ⎥ ⎢ ⎥ ⎣ ⎦j ⎣ 0 ⎦j (18) Since the components of the stiffness vector g . resulting in a generalized Hopf (Neimark) bifurcation. the velocity is substituted with generalized momenta p. If the eigenvalues are inside the unit circle the system is asymptotically stable.τ 2 ) obviously corresponds to a fixed point on the Poincaré section. For the periodic solution. (19) (14) For the the the (15) be a point on the Poincaré section (fixed point). convenience and without loss of generality.τ 1 . the periodic orbit x (s . if at least one of the eigenvalues is outside the circle the system is unstable. June18-21. First. defined inside each of the stage stiffness region: q ′′ + Zq ′ + Π kΩ q = bkΩ + f 0 + f a cos(ητ ) . 0⎥ ⎦ (23) 2 ⎡1 −ω12 ⎤ Π 4Ω = Π 6Ω = ⎢ . Information on the stability of the fixed point can be obtained by studying the eigenvalues of the Jacobian matrix ∂P ∂s . 2007 terms: d j + K j ∆q j = p j . . symmetrybreaking. the iteration procedure needs their computations at each iteration step.e. p(τ j +1 )⎤⎦⎥ denotes the vector of j momenta at the boundary nodes. respectively. − ω21 ⎦ . T while pB = ⎡⎣⎢− p(τ j ) . Piecewise full decoupling method point. real eigenvalues of the Jacobian matrix leave the unit circle through ±1 . (12) (17) d (n ) + K (n ) ∆q (n ) = 0 where: The above equation can be partitioned as: ⎡d ⎤ ⎡K ⎢ B ⎥ + ⎢ BB ⎢ dI ⎥ ⎢ ⎣ ⎦ j ⎣ K IB K BI ⎤ ⎥ K II ⎦⎥ j d (n ) = Aq (n ) + g (q ) + f . T T 2 ⎡ ⎤ b2 Ω = −b8 Ω = ⎣1. The equation (20) can be reformulated as the relationship between the fixed points at the ends of period s2 and s1 which leads to the Jacobian matrix: −1 ⎡ K11 − K12 ∂P ∂s = ⎢ −1 ⎣ K 21 − K 22 K12 K11 −1 ⎤ − K12 . Let: s = [ ∆q . The stability boundary is the unit circle itself. The components of matrix A and vector f are constants and they need to be calculated only once.τ 1 . complex conjugate eigenvalues leave the unit circle away from the real axis.

Results (24) are the vectors of the breakpoints. the system is represented with new local equation of motion. q 1 where z denotes the normal coordinate. 10 1 (26) ˆ . state amplitude i steady (29) q1 . q 1 q1ef 10 0 coincides with the effective amplitude: qef = 2 (q(t ) − qav ) 2 dt T∫ 0 T (28) the response is periodic.78 0. The switch points in each stage stiffness region and the eigenvalues of the state matrix were computed employing the MATLAB routines FZERO and EIG. 2. In the equation (28).25] and f aT = [ 0. effective amplitude (27) ˆ . transient or chaotic motion. The determination of times in which the system changes a stiffness region can be done only numerically. 10 1 i steady state amplitude q1 .6 0. This procedure is well suited [11] because no another reliable way to distinguish the transient and steady state motion at chaotic responses.4 1. effective amplitude q1ef These results are verified with the time finite element method.2 1. l = 1(1)4 is the matrix of eigenvalues and h is the excitation vector. 2007 T 2 2 2 ⎡ −(ω21 + ω22 b3TΩ = −b7 )⎤ Ω = ⎣ (1 + ω12 ). 0. T T 2 T ⎡ 2 ⎤ b5 b4 Ω = −b6 Ω = ⎣ −ω12 .775 0.05 .79 Fig.765 0. 0 ] V.1 . 3. The piecewise full decoupling method is an explicit integration method and the stability of solutions cannot be studied considering the standard stability procedures such as Poincaré map or Floquet theory.77 0.6 1.5. the stability of solutions has to be investigated by studying the periodicity of responses.76 -1 η 0. The responses obtained by explicit metods can be only classified as either periodic or nonperiodic. the piecewise full decoupling computations are performed simulating 128 excitation periods for the single point of excitation frequency. ⎦. ω22 = 1.785 0.767 as shown in figure 3. Besançon (France). 0] are adopted from [9] The matrix A is a real nonsymmetric matrix and its eigenvalues can be calculated using one of numerical routines for the nonsymmetric eigenvalue problem. Each of the local equations of motion (22) has been solved by applying the state-space formulation.8 2 η q1ef Fig. Λ = diag(λl ). June18-21. Therefore. q ′] implies the state matrix A of the form: ⎡ 0 A=⎢ ⎣ − Π kΩ I ⎤ ⎥. T denotes the period of motion while qav is the average of the q(t): 1 qav = ∫ q(t ) dt . − ω22 ⎦ .8 1 1. f 0T = [ 0. Ω = [ 0.6 . ω12 = ω21 = 0. The finite element calculations were performed .25. The dynamical system with clearances starts from an initial position described with one of the local equation of motion. Nonperiodic responses may corespond to a quasiperiodic. −Z ⎦ (25) The piecewise full decoupling method and the time finite element method are applied to determine the bifurcation point of the two-degree-of-freedom dynamical system with two clearances. Obtained eigenvalues enable the similarity transformation which uncouples the equations of motion giving: z′ = Λ z + h where the system with one clearance was studied using the harmonic balance method.4 0. otherwise the response is nonperiodic. The equation (26) has wellknown analytical solutions.12th IFToMM World Congress. T 0 T 10 0. If the steady state amplitude: q= qmax − qmin 2 q1ef10 0 10 -1 10 -2 0 0.2 0. The frequency response in figure 2 indicates the frequency range of the uncorresponding solutions for steady state and effective amplitudes. The separation of the amplitudes (bifurcation point) starts at η = 0. The state T vector y = [ q. Separation of the steady state and effective amplitudes. Frequency response obtained by the piecewise full decoupling method. When the system changes a stage stiffness region. The system parameters ζ 11 = ζ 12 = ζ 21 = ζ 22 = 0. Starting from the trivial initial condition.

2 = −0.6 0. while for the excitation frequency η = 0.772 0.4 = 0.96 (figure 4).62 1.772 0 -0. 2007 with ten four nodes time elements (the third order polynomials are taken as shape functions). Using the previous stable solution as the starting vector only unstable solutions are found. λ1. Frequency response obtained with the previous stable solution as the starting vector. the bifurcation point can be precisely determined.8 2 η q1′ 0. 0.9953 λ3.5 2 2.2 = 1.6 1.5 0.4 1. Phase portraits showing the Neimark bifurcation ˆ1 q 10 0 0.5 q1 10 -1 0. 10 1 λ1.64 0 0.58 1.5 q1′ 10 -2 η = 0. 5. λ3.2 1.773 the eigenvalues yield: q1 Fig.4 = 0. For the excitation frequency η = 0.12th IFToMM World Congress.4 0. June18-21. These results indicate that a complex conjugate pair of the eigenvalues crosses the unit circle away from the real axis resulting in Neimark bifurcation.4 1.4 1. λ1.8 1 1.2 = −0.4 η = 0. the eigenvalues of the Poincarè map are: q1′ 0.2 10 -0.4 1.4256 (31) (unstable solution). 4.5 Fig.773 λ1.6 0.2 0 -0. Considering the response of the linear system as the starting vector.8 2 η Fig. The iteration procedure limited on 100 iterations was accompanied with the stability analysis by the Poincaré map.62 1.2 0.9996i.4 0.0266 ± 0. λ3.6 0.0280 ± 0.4 1 1.8 1 1.6 1.4 0.4 = 0.0008 λ3.58 1.4247i.2 ˆ1 q q1′ 0 η = 0.5 2 2.9939i.0527 ± 0.45 η = 0.2 0. stable solution.45 0.0496 ± 0.64 (asymptotically stable solution).2 1.773 10 -2 0 0. Poincarè sections showing the Neimark bifurcation .77 < η < 0.772 .4267i. Frequency response obtained by the time finite element method (starting vector-response of linear system). stable solution. Besançon (France).4 1 1. the unconverged or unstable solutions are found in the frequency range 0.94 (figure 5). x unstable solution or achieved maximum of 100 iterations 10 1 q1 Fig. 7.4 = 0.55 Imposing small increments in the excitation frequency. 6.6 1. x unstable solution q1 0.55 10 -1 0.6 1.7 < η < 0.4275 (30) 0.2 = 0. but within narrow bounds 0.2 -0.6 0.

June 14-17. Thompson J.T. on Applied Mathematics and Scientific Computing. 185(1): 155-170. Conf.S.785 0. August 1991. Wolf H. 149(1): 91-105. Vrankovic N. and Singh R. and Lau S. Kranjcevic N. and Vrankovic N. and Stegić M. 1986.. The patterns in figures 6. Frequency response of a mechanical system with multiple clearances using the piecewise full decoupling method and the time finite element method. October 1999. Proceedings in Applied Mathematics and Mechanics.78 0.. Stegic M. Nonlinear dynamics and chaos: Geometrical methods for engineers and scientists.. Journal of Sound and Vibration. q 1 q1ef 10 0 [10] η 10 0. 2004. References [1] Chatterjee S. pages 552-556. Stegic M. Stegic M. Journal of Sound and Vibration. 2005. Tianjin. In the finite element calculations. Conclusions [14] The local bifurcation in a dynamical system with clearances is a qualitative change of its behaviour which occurs as the consequence of stiffness discontinuity. 142(1): 49-75. repeated piecewise full decoupling computations (figure 8) agree very well with the results of the time finite element method (30) and (31).76 -1 0. Kranjcevic N. VI. on Computational Plasticity. 191(1): 129-144. Wang Y. In 8th Int. Nonlinear problems in dynamics by the finite element in time method. Barcelona. pages 297-301.M. Zhang W. The piecewise full decoupling method for dynamic problems. Kranjcevic N. 7 confirm the Neimark bifurcation. 3(1): 112-113. June 1992. 2007 The phase portraits and Poincarè sections of the bifurcating solution are determined by using the piecewise full decoupling method.. the bifurcation of the two-degree-of-freedom dynamical system with clearances is studied by using the time finite element method and the piecewise full decoupling method. pages 1146-1149. Periodic forced vibration of unsymmetrical piecewise-linear systems by incremental harmonic balance method. Frequency response obtained by the piecewise full decoupling method (clearances are modelled as trilinear system). 2005.. Meirovitch L. Besançon (France). and Balachandran B.L. Stegic M. [12] i steady state amplitude q1 . on Computational Fluid and Solid Mechanics.K. October 1990. Kranjcevic N. September 5-7. Method for computing steady state response of dynamical systems with clearances. effective amplitude q1ef [13] Modelling the clearances with the above described trilinear system. [3] [4] [5] [6] [7] [8] [9] ˆ . Principles and techniques of vibratons. and Vrankovic N. 1997. Nonlinear dynamics of a spur gear pair. Conf. Journal of Sound and Vibration. August 1995. 155(2): 209-230. 2001. PrenticeHall. June 4-8. In Int. Stegic M. 24(4): 230-237. This disagreement can be understood knowing that the time finite element method cannot solve problems with ideal clearance because the zero stiffness implies the singularity of the tangent matrix K (n ) . 8.B. Dynamics of unsymmetric piecewise-linear/non-linear systems using finite elements in time. Computational Mechanics.12th IFToMM World Congress. Simulation procedure was performed with 4096 excitation periods in transient part regarding the displacements and velocities at the first node of the time finite element mesh as the initial conditions. June18-21. the clearance nonlinearity (3) was approximated by the trilinear system at which the slope of the second stage is 1% of the slope of the first and third stages. and Kranjcevic N. pages 211219. and Vrankovic N. The influence of neglecting small harmonic terms on estimation of dynamical stability of the response of non-linear oscillators. Comparison of the proposed bifurcation analyses shows that both procedures give very accurate results.. Applied nonlinear dynamics. and Ghosh A.775 0. Periodic response of piecewise non-linear oscillators under harmonic excitation. and Stewart H. John Wiley.79 [11] Fig. In this paper. Stability and bifurcation analysis of a two-degree-of-freedom system with clearances. The bifurcation point obtained by the piecewise full decoupling method and the time finite element method is not in agreement. Padmanabhan C. March 1996. and Vrankovic N.77 0. Dubrovnik. .. Journal of Sound and Vibration. John Wiley.765 0. Journal of Sound and Vibration. April 1-4. Mallik A. Nayfeh A. Cambridge. Spectral coupling issues in a two-degree-of-freedom system with clearance non-linearities.H. 1995. 10 1 [2] Kahraman A. In 11th World Congress in Mechanism and Machine Science. December 2003. In 3rd MIT Conf.W. and Singh R. Wong C.

abaqus

abaqus

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd