A three-dimensional nonlinear reduced-order predictive joint model

Yaxin Song, C. J. Hartwigsen, Lawrence A. Bergman, Alexander F. Vakakis

Research output: Contribution to journalArticlepeer-review


Mechanical joints can have significant effects on the dynamics of assembled structures. However, the lack of efficacious predictive dynamic models for joints hinders accurate prediction of their dynamic behavior. The goal of our work is to develop physics-based, reduced-order, finite element models that are capable of replicating the effects of joints on vibrating structures. The authors recently developed the so-called two-dimensional adjusted Iwan beam element (2-D AIBE) to simulate the hysteretic behavior of bolted joints in 2-D beam structures. In this paper, 2-D AIBE is extended to three-dimensional cases by formulating a three-dimensional adjusted Iwan beam element (3-D AIBE). Impulsive loading experiments are applied to a jointed frame structure and a beam structure containing the same joint. The frame is subjected to excitation out of plane so that the joint is under rotation and single axis bending. By assuming that the rotation in the joint is linear elastic, the parameters of the joint associated with bending in the frame are identified from acceleration responses of the jointed beam structure, using a multi-layer feed-forward neural network (MLFF). Numerical simulation is then performed on the frame structure using the identified parameters. The good agreement between the simulated and experimental impulsive acceleration responses of the frame structure validates the efficacy of the presented 3-D AIBE, and indicates that the model can potentially be applied to more complex structural systems with joint parameters identified from a relatively simple structure.

Original languageEnglish (US)
Pages (from-to)59-73
Number of pages15
JournalEarthquake Engineering and Engineering Vibration
Issue number1
StatePublished - Jun 2003


  • Adjusted iwan beam element (AIBE)
  • Bolted joints
  • Multi-layer feed-forward neural networks (MLFF)
  • Nonlinear dynamic analysis
  • Parameter identification

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Building and Construction
  • Geotechnical Engineering and Engineering Geology
  • Mechanical Engineering


Dive into the research topics of 'A three-dimensional nonlinear reduced-order predictive joint model'. Together they form a unique fingerprint.

Cite this