A Bayesian model updating with incomplete complex modal data

A multi‐objective framework for finite element model updating using incomplete modal measurements

“…Besides, Behmanesh and Moaveni 39 and Lam et al 40 applied advanced Markov chain Monte Carlo (MCMC) techniques for probabilistic damage detection in contribution to structural health monitoring (SHM). Also, direct utilization of complex modal data for model updating of damped structure in Bayesian framework has been performed by few researchers like Cheung and Bansal 41 and Das and Debnath 42 . Few other useful works may be worth mentioning here, for example, model updating using multi‐resolution Bayesian non‐parametric general regression by Yuen and Ortiz, 43 sensitivity clustering based Bayesian updating by Jang and Smith, 44 Bayesian model updating by utilizing modal data from multiple setups by Zhang et al, 45 and so forth.…”

Computational inference of vibratory system with incomplete modal information using parallel, interactive and adaptive Markov chains

“…Finite element (FE) method nowadays is pervasive in the analysis of mechanical structures [1][2][3][4]. A critical step in FE analysis is model updating in which certain modeling parameters are identified/updated to minimize the discrepancy between FE-based response prediction and the actual measurement.…”

“…A critical step in FE analysis is model updating in which certain modeling parameters are identified/updated to minimize the discrepancy between FE-based response prediction and the actual measurement. Indeed, in recent years, there have been growing interests in FE model updating using vibration response data to facilitate a series of inter-related applications, such as damage identification, design optimization, and structural control, etc [2][3][4]. In dynamic systems, FE model updating generally is conducted by employing the responses in either time or frequency domain [5][6][7].…”

“…Bayesian inference not only can avoid the direct inversion for parameter estimation required in some sensitivity-based methods that may introduce some numerical issues, but also can directly incorporate various sources of uncertainties into model updating procedure [14]. Owing to its intrinsic advantages, there have been considerable successes in utilizing Bayesian inference to solve a variety of engineering problems [4,[15][16][17][18]. Additionally, as the training scheme of meta-models, which are essentially implicit statistical regression models, is generally established upon Bayesian inference, exploring the potential close-form of optimization objective function under Bayesian framework may benefit meta-model training efficiency [19].…”

“…Assuming that we divide Q bins within the range [0, 1] for all 4 model parameters, the total number of bins will be 4 Q . When choosing a large value of Q to reflect the continuous property of model parameters, the sample count procedure will be quite costly in order to loop over all 4 Q bins. To facilitate this counting procedure, in this research we set a small number of Q as 20.…”