Research Papers

A Methodology for the Modeling of Forced Dynamical Systems From Time Series Measurements Using Time-Delay Neural Networks

[+] Author and Article Information
John Zolock

 Exponent Failure Analysis Associates, Natick, MA 01760

Robert Greif

Department of Mechanical Engineering, Tufts University, Medford, MA 02155

J. Vib. Acoust 131(1), 011003 (Dec 29, 2008) (10 pages) doi:10.1115/1.2981096 History: Received February 01, 2006; Revised May 30, 2008; Published December 29, 2008

The main goal of this research was to develop and present a general, efficient, mathematical, and theoretical based methodology to model nonlinear forced-vibrating mechanical systems from time series measurements. A system identification modeling methodology for forced dynamical systems is presented based on a dynamic system theory and a nonlinear time series analysis that employ phase space reconstruction (delay vector embedding) in modeling dynamical systems from time series data using time-delay neural networks. The first part of this work details the modeling methodology, including background on dynamic systems, phase space reconstruction, and neural networks. In the second part of this work, the methodology is evaluated based on its ability to model selected analytical lumped-parameter forced-vibrating dynamic systems, including an example of a linear system predicting lumped mass displacement subjected to a displacement forcing function. The work discusses the application to nonlinear systems, multiple degree of freedom systems, and multiple input systems. The methodology is further evaluated on its ability to model an analytical passenger rail car predicting vertical wheel∕rail force using a measured vertical rail profile as the input function. Studying the neural modeling methodology using analytical systems shows the clearest observations from results, providing prospective users of this tool an understanding of the expectations and limitations of the modeling methodology.

Copyright © 2009 by American Society of Mechanical Engineers
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Figure 1

General architecture for a time-delay neural network (TDNN)

Grahic Jump Location
Figure 2

Examples of the analytical systems studied

Grahic Jump Location
Figure 3

Vertical rail track geometry

Grahic Jump Location
Figure 4

Example segment of input-output data

Grahic Jump Location
Figure 5

(a) Mutual information and (b) percentage of false nearest neighbors

Grahic Jump Location
Figure 6

Performance measure for training and validation

Grahic Jump Location
Figure 7

TDNN validation output over (a) actual measured system response and (b) residual.

Grahic Jump Location
Figure 8

Exploded and sectioned side front views of an equalizer beam truck

Grahic Jump Location
Figure 9

Segment of input-output data

Grahic Jump Location
Figure 10

(a) Mutual information and (b) percentage of false nearest neighbors

Grahic Jump Location
Figure 11

Mean absolute error performance measure

Grahic Jump Location
Figure 12

Comparison of (a) predicted and actual and (b) residual from validation

Grahic Jump Location
Figure 13

Analytically generated white noise dataset

Grahic Jump Location
Figure 14

(a) Plots of mutual information and (b) false nearest neighbors




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In