Bayesian model updating of nonlinear systems using nonlinear normal modes

Abstract

This paper presents a Bayesian model updating methodology for dynamical systems with geometric nonlinearities based on their nonlinear normal modes (NNMs) extracted from broadband vibration data. Model parameters are calibrated by minimizing selected metrics between identified and model-predicted NNMs. In the first approach, a deterministic formulation is adopted, and parameters are updated by minimizing a nonlinear least-squares objective function. A probabilistic approach based on Bayesian inference is next investigated, where a Transitional Markov Chain Monte Carlo is implemented to sample the joint posterior probability distribution of the nonlinear model parameters. Bayesian model calibration has the advantage to quantify parameter uncertainty and to provide an estimation of model evidence for model class selection. The two formulations are evaluated when applied to a numerical cantilever beam with geometrical nonlinearity. The NNMs of the beam are derived from simulated broadband data through nonlinear subspace identification and numerical continuation. Accuracy of model updating results is studied with respect to the level of measurement noise, the number of available datasets, and modeling errors.

Accounting for Modeling Errors and Inherent Structural Variability through a Hierarchical Bayesian Model Updating Approach: An Overview

“…The response can then be predicted probabilistically using the updated model, by propagating the parameters uncertainty. Several applications of Bayesian model updating on numerical [ 19 , 24 , 25 , 26 , 27 , 28 , 29 , 30 ], laboratory [ 31 ] and real-world [ 32 , 33 , 34 , 35 ] structures can be found in the literature.…”

Structural dynamic nonlinear model and parameter identification based on the stiffness and damping marginal curves

“…With the deepening of study, the nonlinear system identification methods have made significant progress in recent years, such as the probabilistic framework based on the well‐known Bayesian theorem, the widely used homogeneity method, and the frequency response function (FRFs)‐based method . In general, the nonlinear system identification methods can be classified as time domain, frequency domain, and time–frequency domain methods.…”

Modeling Error Estimation and Response Prediction of a 10-Story Building Model Through a Hierarchical Bayesian Model Updating Framework

“…The classical Bayesian model updating framework cannot explicitly quantify the modeling errors since all three sources of uncertainty mentioned above are lumped into one term. However, the classical formulation is useful for model class selection among competing model forms (Ching and Chen, 2007;Song et al, 2018). Error-domain model falsification algorithm is shown to be capable of falsifying model instances/classes in the view of compatibility with measurement by avoiding assumptions on the exact distribution of modeling errors and residual dependency Pasquier and Smith, 2015).…”