Reduced order modeling of hysteretic structural response and applications to seismic risk assessment

• Published in: Engineering structures Vol. 209; p. 110135
• Main Authors: Patsialis, D., Taflanidis, A.A.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 15.04.2020; Elsevier BV
• Subjects: Accuracy; Calibration; Computer applications; Dynamic structural analysis; Earthquake engineering; Earthquakes; Environmental risk; Equations of motion; Excitation; Finite element method; Hysteresis; Hysteretic structural response; Linear equations; Mathematical models; Matrix methods; Reduced order modeling; Reduced order models; Risk assessment; Seismic activity; Seismic engineering; Seismic hazard; Seismic response; Seismic risk assessment; Stiffness matrix; Structural response; Time-history based optimization; Seismic risk assessment; Time-history based optimization; Reduced order modeling; Hysteretic structural response
• ISSN: 0141-0296; 1873-7323
• DOI: 10.1016/j.engstruct.2019.110135

•Development of reduced order model (ROMs) for hysteretic structural response is examined.•Linear ROM is developed through static condensation of the original high-fidelity FEM.•Nonlinear ROM is subsequently formulated by replacing linear stiffness components with hysteretic ones.•Calibration is established through comparison of the time-history response to the initial FEM.•Applications to seismic loss assessment for different structures illustrate significant computational benefits. A reduced order modeling approach is presented to alleviate the computational burden associated with using high-fidelity finite element models (FEMs) to describe hysteretic structural response for earthquake engineering time-history analysis. The reduced order model (ROM) is developed using data from the original high-fidelity FEM. Static condensation is first used to obtain the condensed stiffness matrix and the linear equations of motion for the dynamic degrees of freedom (DoFs). The restoring forces prescribed by the linear stiffness matrix are then substituted with hysteretic ones by replacing the linear springs connecting each of the DoFs with hysteretic ones. Different hysteretic models are considered, including peak-oriented, Masing and Bouc-Wen type of hysteresis, whereas more complex relationships obtained by combining multiple simpler hysteretic models are also discussed. The hysteretic spring parameters are calibrated by comparing the reduced order model time-history response to the time-history response of the initial FEM for a range of different excitations. The characteristics for each of the considered springs are separately selected and an efficient solution of the associated calibration problem is facilitated through a sequential, hierarchical approach. The excitations utilized for the reduced order model calibration are carefully selected so that nonlinear characteristics of the FEM are appropriately excited to support the tuning of all the important hysteretic spring features. The accuracy and the computational savings of the calibrated reduced order model are evaluated for risk assessment applications, separately examining different levels of intensity. Comparison extends to three structures, with the high-fidelity FEMs developed in OpenSees.