Nonlinear dynamic physical systems exhibit a rich variety of behaviors. In many cases,
the system response is unstable, and the behavior may become unpredictable. Since
an unstable or unpredictable response is usually undesirable in engineering practice,
the stability characterization of a system's behavior becomes essential.
In this work, a numerical procedure to characterize the dynamic stability of con-
tinuous solid media, discretized using finite element methods, is proposed. The pro-
cedure is based on the calculation of the maximum Lyapunov characteristic exponent
(LCE), which provides information about the asymptotic stability of the system re-
sponse. The LCE is a measure of the average divergence or convergence of nearby
trajectories in the system phase space, and a positive LCE indicates that the sys-
tem asymptotic behavior is chaotic, or, in other words, asymptotically dynamically
unstable. In addition, a local temporal stability indicator is proposed to reveal the
presence of local dynamic instabilities in the response. Using the local stability indi-
cator, dynamic instabilities can be captured shortly after they occur in a numerical
calculation. The indicator can be obtained from the successive approximations of the
response LCE calculated at each discretized time step. Both procedures can also be
applied to fluid-structure interaction problems in which the analysis focuses on the
behavior of the structural part.
The response of illustrative structural systems and fluid flow-structure interac-
tion systems, in which the fluid is modeled using the Navier-Stokes equations, was
calculated. The systems considered present both stable and unstable behaviors, and
their LCEs and local stability indicators were computed using the proposed proce-
dures. The stability of the complex behaviors exhibited by the problems considered
was properly captured by both approaches, confirming the validity of the procedures
proposed in this work.
Thesis Supervisor: Klaus-Jiirgen Bathe
Title: Professor of Mechanical Engineeing
Author(s): Sandra Rugonyi
Publisher: MIT
Year: 2001
Language: English
Pages: 125