Bivariate Gaussian mixture-based equivalent linearization method for stochastic seismic analysis of nonlinear structures

摘要

To address challenges in stochastic seismic analysis of nonlinear structures, this paper further develops a recently proposed Gaussian mixture-based equivalent linearization method (GM-ELM). The GM-ELM uses a Gaussian mixture distribution model to approximate the probabilistic distribution of a nonlinear system response. Using properties of the Gaussian mixture model, GM-ELM can decompose the non-Gaussian response of a nonlinear system into multiple Gaussian responses of linear single-degree of freedom oscillators. With the set of the equivalent linear systems identified by GM-ELM, response statistics as crossing rate and first-passage probability can be computed conveniently using theories of linear random vibration analysis. However, the original version of GM-ELM may lead to an inaccurate estimate because of the heuristic parameters of the linear system introduced to supplement insufficient information. To overcome this limitation and define unique equivalent linear systems, this paper proposes a further developed version of GM-ELM, which uses a mixture of bivariate Gaussian densities instead of univariate models. Moreover, to facilitate the use of elastic response spectra for estimating the mean peak responses of a nonlinear structure, a new response spectrum combination rule is proposed for GM-ELM. Two numerical examples of hysteretic structural systems are presented in this paper to illustrate the application of the bivariate GM-ELM to nonlinear stochastic seismic analysis. The analysis results obtained by the bivariate GM-ELM are compared with those obtained by the univariate GM-ELM, the conventional equivalent linearization method, the tail equivalent linearization method, and Monte Carlo simulation. The supporting source code and data are available for download at https://github.com/yisangri/GitHub-bGM-ELM-code.git