A parametric reduced-order model using substructural mode selections and interpolation. (February 2019)

• Record Type: Journal Article
• Authors: Lee, Jaehun
• Is Part Of: Computers & structures. Volume 212(2019)
• Page Start: 199
• Page End: 214
• Publication Date: 2019-02
• Subjects: Parametric reduced-order model -- Component mode synthesis -- Substructural mode selection -- Interpolation -- Congruence transformation; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2018.10.018
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: An efficient, parametric reduced-order model is proposed for large-scale dynamical systems. The problem of crossing between dominant and residual modes is resolved by a new algorithm for the intermediate model. The substructural mode selection criterion was employed, maintaining online capability of the PROM. A high-dimensional parametric input space is efficiently handled by the proposed method. Abstract: This paper presents a parametric reduced-order model (PROM) combined with component mode synthesis considering the interpolation and selection of substructural modes. For the interpolation-based PROM, mode crossing occurs depending on the change of parameter values. This problem is resolved by applying a congruence transformation to each mode. However, if there is a mode crossing between dominant and residual modes due to the mode selection following the frequency cut-off method, the interpolation of dominant substructural modes generates erroneous results, since there are missing modes at some operating points. In this study, we propose (i) to introduce an intermediate model in the offline stage of the ROM construction to decrease the interpolation error, and (ii) to select proper substructural modes by applying a mode selection criterion for the enhancement of the ROM accuracy, which is realized under the framework of dynamic substructuring. Since the intermediate model is used in the offline stage, there is little loss of efficiency in the online stage. Consequently, the characteristics of interpolation-based PROM is maintained, realizing a near real-time adaptation of the ROM. Various numerical examples, including dynamic response optimization with a high-dimensional parametric input space, are investigated to evaluate the performance of the proposed method.