A model updating method based on response surface models of reserved singular values. (October 2018)

• Record Type: Journal Article
• Authors: Zhang, Yong; Hou, Zhi-chao
• Is Part Of: Mechanical systems and signal processing. Volume 111(2018)
• Page Start: 119
• Page End: 134
• Publication Date: 2018-10
• Subjects: Model updating -- Response surface -- Singular value decomposition -- Support vector machine -- Impulse response function; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2018.03.038
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: Impulse response functions are derived from frequency response functions to construct the phase space matrices. Dominated singular values of the state space matrix are chosen to represent the characteristics of the target structure. With parameters to be updated as the input, a support vector machine based response surface model is established and validated for each of the reserved singular value. A minimization problem is defined and solved by a genetic algorithm for the accurate values of the parameters to be updated. We verified the proposed method using calculated FRFs or measured FRFs, respectively. Abstract: A method for finite element model updating is put forward using response surface models of the truncated singular values with respect to model parameters to be updated. The impulse response function is firstly derived in time series from the known frequency response functions (FRFs), and a phase space matrix is then constructed through a time-delay method. Dominated singular values of the state space matrix are chosen to represent the characteristics of the frequency response functions and the system behind. A support vector machine based response surface model is established and validated for each of the reserved singular value with respect to the updating parameters. Taking the error between the reserved singular values calculated by the response surface models and those from the known FRFs as the cost function, a minimization problem is framed, which is then solved by a genetic algorithm for the true values of the updating parameters. Two examples about vehicle chassis components, with calculated or measured FRFs, are given to verify the applicability and efficiency of the proposed method.